Planet Code4Lib

Journal of Web Librarianship: A Streamlined AI-Powered Workflow for Enriching Library Newspaper Archives

2026-05-07T03:20:32+00:00

Coral Sheldon-Hess: AI policies and informed consent

2026-05-07T01:32:31+00:00

There sure is a lot of noise out there about so-called “artificial intelligence.” (I won’t bother with the scare quotes through the whole post, but it’s worth noting: “AI” is a term that’s been in use, mostly for marketing, since the 1950s (here is the term’s original usage), and as such, “resists definition because it is continually reappropriated by people to mean different things.”) I’m a little sorry to add to it, but there are many people out there being pressured to use AI products, who I think should be informed of the risks and allowed to decide whether to use them or not. My primary focus, here, will be “generative AI,” the chatbots that use huge corpora of text and images to generate outputs that appear novel, and my approach will be largely practical; I’ll save ethical and environmental concerns for the last section, and mostly link to better sources than this post. I also limit myself to work applications — so I removed a section about chatbots’ effects on mental health, despite my belief that their sycophancy could potentially cause problems at work.

Contents of this post

To save you some scrolling in case you’re here for something specific.

And last, but in a just world, not least:

Ethics and the environment (mostly a list of resources)

The benefits are usually negligible, but often nonexistent and sometimes even negative

Let’s cover this first, since it’s the part management is most likely to care about—and because it’s counter to everything the marketers are telling us, I have citations—AI does not improve individual performance in any meaningful way. Occasionally, for some tasks, there are small improvements, but when weighed against the near certainty that these products will increase in price, the potential for catastrophic mistakes, and the likelihood of damage to the individuals using these products (which is the bulk of this post), I argue that management has no incentive to force AI use on workers.

Now

According to MIT’s The State of AI in Business in 2025 (emphasis mine), “Despite $30-40 billion in enterprise investment into GenAI, this report uncovers a surprising result in that 95% of organizations are getting zero return.”

An uncomfortable number of publications about worker productivity focus on qualitative assessments:

From the Wall Street Journal, “Two-thirds of nonmanagement staffers said they saved less than two hours a week or no time at all with AI. More than 40% of executives, in contrast, said the technology saved them more than eight hours of work a week.”
Also WSJ, summarizing a different study: “A new report from the business-software company Workday goes so far as to call frustrations with the technology an ‘AI tax’ on productivity. Though 85% of the roughly 1,600 employees it surveyed reported saving one to seven hours a week by using AI, much of the time was offset by having to correct errors and rework AI-generated content.”
From upwork, “Nearly half (47%) of employees using AI say they have no idea how to achieve the productivity gains their employers expect, and 77% say these tools have actually decreased their productivity and added to their workload.” (The same executive summary says 71% of workers are burnt out, which… yeah, that sounds right.)

On the quantitative front:

For programmers, we have a measured 19% slow-down in development speed in early 2025. Notably, the developers thought they had improved by 20%, which suggests self-reported gains from AI should be considered optimistic, at best.
Uplevel (Uplevel’s report download page, or here’s a summary if you aren’t prepared to share an email address) studied 800 programmers with and without Copilot assistance and found that, while their speed did not change, programmers using AI pushed 41% more bugs.
“A UK government department’s three-month trial of Microsoft’s M365 Copilot has revealed no discernible gain in productivity – speeding up some tasks yet making others slower due to lower quality outputs.” From The Register.
What it says on the tin: AI can cost more than workers now, by Madison Mills, April 2026.

The future

I can almost hear the AI fans shouting, “But that’s 2025! We’re looking at the future!” And, you know what? That’s fair. Let’s talk about the future.

A great deal of the hype is based on a simple bet: that these products, which are currently pretty mid, will improve over time. But it’s a bad bet. Last September researchers at OpenAI proved that there is no way to create a large language model that doesn’t hallucinate. (I linked you to a summary, there, but if you like mathematical proofs, here is the original OpenAI article.) These are inherently probabilistic technologies, which means their output cannot be constrained to what we commonly refer to as “the truth.” No amount of engineering, scaling, or guardrails can prevent these machines from lying to us. We’re also running out of scaling opportunities; the internet, though large, is finite. A second bet, that the LLMs can feed each other, leading to a “superintelligence,” has also been shot down by mathematical proof. There’s a convergence point beyond which they don’t improve and actually get significantly worse. (I’ll be honest: mathematical proofs aren’t my thing. This post is not at all restrained in its rhetoric, but it does a good job explaining what’s going on in the paper.)

If a singularity is coming, it won’t be from this branch of technology.

So, OK, genAI won’t really get better over time, unless we pour a whole lot more human-generated data into it; and even then, there are limits. The companies can make little refinements, but we’re stuck with hallucinations and they need to stop slurping up everything, lest their products degrade further in quality.

Conclusion: there is no benefit to the enterprise in pushing individual contributors to use AI.

(Question: do I need a follow-up post about how silly the idea of replacing workers with AI is? Klarna may be one of the 55% of businesses that admit they shouldn’t have cut workforce in place of AI — and that’s just the ones willing to admit to being wrong. Even for replacing freelancers, AI’s a no-go.)

Harms to individual users

I’d like to believe management would care equally about this, and to be fair, a lot of managers will: good managers want their employees to be successful, even after they’ve moved on. Above a certain level, though, it seems like individuals stop mattering. So at this point, I’m talking mostly to line managers and to individuals who need to decide whether to implement genAI products in their own work.

Not all of these harms are equally applicable to everybody, but I believe they should all be shared with anyone who is thinking of using these products for their work.

AI usage damages critical thinking and encourages cognitive surrender

“While LLMs offer immediate convenience, our findings highlight potential cognitive costs. Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels.” – Your brain on ChatGPT: Accumulation of cognitive debt when using an ai assistant for essay writing task, by Kosmyna et. al., arXiv preprint, 2025.
Critical thinking is inversely correlated with trust in AI systems. – The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers, by Lee et. al., Proceedings of the CHI Conference on Human Factors in Computing Systems, April 2025.
“Across studies, participants with higher trust in AI and lower need for cognition and fluid intelligence showed greater [cognitive] surrender.” – Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender, by Steven D. Shaw and Gideon Nave, The Wharton School Research Paper, February 2026.
This interview’s summary of the next paper is good: “all users show a significant inability to assess their performance accurately when using ChatGPT. In fact, across the board, people overestimated their performance. … ‘We found that when it comes to AI, the Dunning-Kruger Effect vanishes. In fact, what’s really surprising is that higher AI literacy brings more overconfidence,’ says Professor Robin Welsch. ‘We would expect people who are AI literate to not only be a bit better at interacting with AI systems, but also at judging their performance with those systems – but this was not the case.’” – AI makes you smarter but none the wiser: The disconnect between performance and metacognition, by Fernandes et. al., Computers in Human Behavior, February 2026.

AI products encourage dependence and decrease self-confidence

This one’s interesting because this person is pro-AI. But I found this quote telling: “If you think about it, that 30-second wait for AI responses can be seen as a variable ratio schedule — Random rewards delivered at unpredictable intervals — the same psychological pattern that makes slot machines, social media, and mobile games addictive.” – Vibe Coding Is Creating Braindead Coders, by Namanyay Goel, Blog post, September 2025
This one will hurt to read, a bit, if you’re already using these products, and I’m sorry. It’s a good warning to the folks who aren’t, yet, and it comes with citations. Generative AI runs on gambling addiction — just one more prompt, bro!, by David Gerard, June 2025.
“The study included 1,923 online adult participants from the United States and Canada who were told to use commercially available AI programs to complete 10 simulated work tasks, such as developing plans with incomplete or evolving information, interpreting ambiguous data, and articulating reasoning for strategic decisions.

“After the tasks, 58% of the participants agreed that AI ‘did most of the thinking’ to complete the work, especially in activities related to planning or sequencing. Those participants also reported reduced confidence in their own independent reasoning, lesser perceived ownership of ideas, and making trade-offs between task speed and depth of thought.” (quote from this summary) Furthermore, “Greater prompt dependence and lower override frequency were associated with reduced self-reported confidence in independent reasoning” – Generative Artificial Intelligence Reliance and Executive Function Attenuation: Behavioral Evidence of Cognitive Offload in High-Use Adults, by Sarah Baldeo in Technology, Mind, and Behavior, 2026.

AI usage hinders skill formation and learning on novel tasks

“We find that AI use impairs conceptual understanding, code reading, and debugging abilities, without delivering significant efficiency gains on average.” – How AI Impacts Skill Formation, by Judy Hanwen Shen and Alex Tamkin, January 2026.
LLM summaries produce shallower learning and understanding than a traditional web search. – Experimental evidence of the effects of large language models versus web search on depth of learning, by Shiri Melumad and Jin Ho Yun, October 2025.
“The analysis of these examples suggests that automation can result in the loss of expertise due to reduced opportunities for learning from deliberate practice and experienced colleagues, and from working on progressively more complex tasks.” – The Impact of Artificial Intelligence on Expertise Development: Implications for HRD, by Alexandre Ardichvili, Advances in Developing Human Resources, 2022.

AI usage erodes expertise and damages performance even on familiar tasks

Using AI assistance hurts performance on both arithmetic and reading tasks as soon as the AI is removed. – AI Assistance Reduces Persistence and Hurts Independent Performance (website, arXiv preprint), by Liu et. al., arXiv preprint, April 2026.
“[O]ver the course of six months, clinicians became over-reliant on AI recommendations and became themselves ‘less motivated, less focused, and less responsible when making cognitive decisions without AI assistance.'” – New Study Suggests Using AI Made Doctors Less Skilled at Spotting Cancer, by Miranda Jeyaretnam, Time, August 2025 (and here is the study it references)

AI usage amplifies individuals’ biases and flaws in judgment, in ways that are invisible to them

“AI systems can exhibit biased judgements in domains ranging from perception to emotion. Here, in a series of experiments (n = 1,401 participants), we reveal a feedback loop where human–AI interactions alter processes underlying human perceptual, emotional and social judgements, subsequently amplifying biases in humans. This amplification is significantly greater than that observed in interactions between humans, due to both the tendency of AI systems to amplify biases and the way humans perceive AI systems. Participants are often unaware of the extent of the AI’s influence, rendering them more susceptible to it.” – How human–AI feedback loops alter human perceptual, emotional and social judgements, by Moshe Glickman and Tali Sharot, Nature Human Behaviour, volume 9, pages 345–359 (2025).
Autocomplete affects the answers people give, and most concerningly, “the people in the study didn’t tend to think the AI autocomplete suggestions were biased or to notice that they had changed their own thinking on an issue in the course of the study.” – AI autocomplete doesn’t just change how you write. It changes how you think, by Claire Cameron, Scientific American, March 2026, extending a 2023 study by Jakesh et. al.
The bullets above are a problem because these models are demonstrably biased.
- “Here, we advance studies of generative language model bias by considering a broader set of natural use cases via open-ended prompting… In this setting, we find that across 500,000 observations, generated outputs from the base models of five publicly available language models (ChatGPT 3.5, ChatGPT 4, Claude 2.0, Llama 2, and PaLM 2) are more likely to omit characters with minoritized race, gender, and/or sexual orientation identities compared to reported levels in the U.S. Census, or relegate them to subordinated roles as opposed to dominant ones. We also document patterns of stereotyping across language model–generated outputs with the potential to disproportionately affect minoritized individuals.” – Intersectional biases in narratives produced by open-ended prompting of generative language models, by Shieh et. al., Nature Communication, 2026.
- “Essays attributed to Black students received more praise and encouragement, sometimes emphasizing leadership or power. … Essays labeled as written by Hispanic students or English learners were more likely to trigger corrections about grammar and ‘proper’ English. When the student was identified as white, the feedback more often focused on argument structure, evidence and clarity — the kinds of comments that can push writers to strengthen their ideas. The AI models addressed female students more affectionately and used more first-person pronouns.” AI gives more praise, less criticism to Black students: Identical essays get different feedback in Stanford study, by Jill Barshay, April 2026; original article preprint.
- “Both large language models significantly underestimated disability in a population of people, and linguistic analysis showed that descriptions of people, patients, and athletes with a disability were generated as having significantly fewer favorable qualities and significantly more limitations than people without a disability in both ChatGPT and Gemini.” – Disability Ethics and Education in the Age of Artificial Intelligence: Identifying Ability Bias in ChatGPT and Gemini, by Urbina et. al., Archives of Physical Medicine and Rehabilitation, January 2025.

Sending uncorrected “workslop” to someone else is always inappropriate

Executive function theft (digression: this is such a great post; it really clarified something I’d observed but hadn’t previously had a cohesive description for) has been a problem in workplaces since time immemorial, but we’re already seeing genAI products making it worse, with huge productivity costs to workers receiving “workslop” products. (“Workslop” definition: “AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task.”) As Harvard Business Review summarizes the issue: “When coworkers receive workslop, they are often required to
take on the burden of decoding the content, inferring missed or false context. A cascade of effortful and complex decision-making processes may follow, including rework and uncomfortable exchanges with colleagues.”

I have to wonder if this is some of the time savings reported by executives: they do a little prompting, send off the output, and one of their employees is stuck doing the difficult part. Anyway…

Besides the obvious, “let people opt out,” a second concrete proposal I made for my workplace’s AI Strategy was “AI products should not be used to generate outputs that can’t be tested and corrected by the team using the product.” (Even that is too broad; they should be tested and corrected by the person using the product.)

An example of an acceptable use: someone needs a ton of data parsed, and Excel is choking on the volume; they use a genAI product, which creates a pandas script to parse the data for them. The code itself isn’t the point and likely won’t be used beyond that session. They spot check the output of the script with smaller portions of their Excel file, to make sure it isn’t totally off base, before using it to make decisions.
An example of an unacceptable use: someone uses a genAI product to create an application and asks the technology group to deploy it. The technology group, if they’re doing due diligence, now has to read through and probably correct that code. (So this person has just generated a ton of work for another team, when instead they should have worked with the experts on that team, who likely would have been able to code something simpler, or at least something they understand well enough to deploy safely.)
Acceptable: someone uses a genAI product to generate a bulleted list of some sort, within their own knowledge domain. They go through, add necessary context, remove duplicates and fix weird phrasing. When it’s clean, they send it off to their task force, to add to a final report.
Unacceptable: someone uses a genAI product to generate a bulleted list about something they don’t fully understand. They send it off to their task force uncorrected, and someone else will now have to do the work to fix it.

AI outputs can be hard to parse

I genuinely don’t know how universal this is. I can find a lot of neurodivergent folks talking about the difficulties they have with AI output, so I know at least a subset of people have a really difficult time with this. (Of course, others are capitalizing on it; but even there, at least some of what’s happening is full rewrites.)

I will say, for myself, that I do not have a diagnosis of autism or ADHD, and I slide right off AI-generated text. Have you ever been reading a book, or you thought you were, and you realize you have no idea what the last page said? Reading AI writing is always like that for me; if it’s something I’m going to be in any way responsible for, I consistently have to close the file and write my own, rather than try to edit it.

I’ll link you to some other folks’ reports of their experiences with it, as well:

Bluesky thread that describes the problem (with, sorry, a good deal of profanity and nothing positive to say about genAI)
Single Bluesky post about the difficulty of double-checking flawed output, when one already questions themselves (includes unkind words about AI boosters)
Bsky post about the difficulty of reading LLM output for someone with ADHD, plus two more posts saying the same thing

AI outputs must be understood to be corrected (and they always need to be corrected)

“Before we can safely change code, we first need to understand it – understand what it does, and also oftentimes why it does it the way it does. In that sense, this is nothing new.

“What is new is the scale of the problem being created as lightning-speed code generators spew reams of unread code into millions of projects.

“Teams that care about quality will take the time to review and understand (and more often than not, rework) LLM-generated code before it makes it into the repo. This slows things down, to the extent that any time saved using the LLM coding assistant is often canceled out by the downstream effort.” – Comprehension Debt: The Ticking Time Bomb of LLM-Generated Code, by Jason Gorman, September 2025.
Companies are drowning in code, without enough people to read, understand, and review it. The big bang: a.i. has created a code overload, by Mike Isaac and Erin Griffith, April 2026.

Ethics and the environment

I pushed this part to the bottom, despite it being so much a part of my original decision to become an AI vegetarian, because every organization creating an AI policy (unless that policy can be summarized as “don’t”) will already have determined that these concerns are secondary to their other goals. Organizations might give lip service to “using AI ethically,” but the people whose opinions I most respect on this topic generally argue that, for people empowered to make organizational decisions, there is no ethical use of generative AI products. (Well, OK, a small subset of the people I most respect are arguing that, perhaps, if you use small models on standard consumer hardware, maybe it’s not so bad — again, they’re arguing from an ethical and environmental standpoint, and many of the harms outlined above still apply. I actually have a use for Whisper, though you’ll note I spent all this time writing this post, rather than deploying it on my Mac.)

At a minimum, I think that if you have the money, you should spend the $20 (unless you can find a free screening near you) and just under 2 hours to watch Ghost in the Machine. It does a pretty good job describing the eugenicist aspects of AI, and, crucially, it spends time with data workers in the Global South whose work makes these products possible. It’s less gut-wrenching than reading some of their first-hand accounts, but it’s enough that you can no longer pretend to think these products aren’t hurting anyone, just by existing.

Some other folks who have done a better job covering the social and environmental costs than I ever could:

Mulvaney, Katy. (2026). Don’t use AI. Southern Connecticut State University.
Guest, O., Suarez, M., Müller, B., van Meerkerk, E., Oude Groote Beverborg, A., de Haan, R., Reyes Elizondo, A., Blokpoel, M., Scharfenberg, N., Kleinherenbrink, A., Camerino, I., Woensdregt, M., Monett, D., Brown, J., Avraamidou, L., Alenda-Demoutiez, J., Hermans, F., & van Rooij, I. (2025). Against the Uncritical Adoption of ‘AI’ Technologies in Academia. Zenodo.
Slater, K. (2025). Against AI: Critical Refusal in the Library. Library Trends 73(4), 588-608.

I want to shout out to people, books, and articles that helped organize and clarify my thinking, even if I didn’t cite them explicitly here

(people) The AI Skeptics Reading Group, which you can join too if you’re interested.
(article) Baer, Andrea. (2025). Investigating the “Feeling Rules” of Generative AI and Imagining Alternative Futures, published in In the Library with the Lead Pipe
(book) Bender and Hanna. (2025). The AI Con.
(book) Eubanks, Virginia. (2018). Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor. (yes, this is about traditional ML, but it’s still relevant)
(book) McQuillan, Dan. (2022). Resisting AI: An Anti-fascist Approach to Artificial Intelligence.
(article) Moser, Anthony. (2025). I am an AI hater.
(book) Madhumita, Murgia. (2024). Code Dependent: Living in the Shadow of AI.
(article) Potter, Ned. (2026). What do we know about GenAI?
(speech, transcribed) Watters, Audrey. (2025). AI Grief Observed.

And why not share my TBR (to-be-read) list, while we’re here

Williams, Rua. (2025). Disabling Intelligences: Legacies of Eugenics and How We are Wrong about AI.
Hao, Karen. (2025). Empire of AI: Dreams and nightmares in Sam Altman’s OpenAI.
Merchant, Brian. (2023). Blood in the Machine: The Origins of the Rebellion Against Big Tech.
Benjamin, Ruha. (2019). Race After Technology: Abolitionist Tools for the New Jim Code.

A final note:

I’m ready to hit “publish,” but at the same time, this does not quite feel finished. I reserve the right to come back and add citations (and you should feel comfortable sending me any big ones I missed, of course!) and, if needed, entire sections. I apologize if that messes up your RSS reader or is otherwise frustrating for you.

Xe Iaso: Maybe you shouldn't install new software for a bit

2026-05-07T00:00:00+00:00

In the wake of copy.fail, there are more vulnerabilities that have been announced:

Right now would be one of the best times for a supply chain attack via NPM to hit hard.

Outside of Linux kernel patches from your distro, I think it's probably a good idea to put a moratorium on installing new software for a week or so.

Open Knowledge Foundation: The Neurotech Booth at RightsCon That Never Happened

2026-05-06T19:07:29+00:00

What testing commercial neurotech has taught us, and why these technologies require informed engagement now.

The post The Neurotech Booth at RightsCon That Never Happened first appeared on Open Knowledge Blog.

HangingTogether: No edge case: Understanding AI opportunities through Arabic metadata workflows

2026-05-06T18:32:12+00:00

We are grateful to our colleague and guest author Omar Farhoud, Sales Manager in the EMEA region, for sharing his perspective on how AI might benefit Arabic-language metadata workflows, as well as risks and limitations to be aware of. For more on implementing AI in metadata workflows, please see our blog series on the topic.

Photo by Nahrizul Kadri on Unsplash

Recent discussions on the OCLC Research blog have explored how artificial intelligence (AI) is beginning to reshape cataloging and metadata workflows, particularly in addressing backlogs and improving efficiency. Conversations like these, however, are often grounded in environments where English-language metadata dominates. Bringing in a Middle Eastern perspective, especially from Arabic-speaking libraries, introduces a different set of conditions shaped by multilingual practice, script diversity, and issues of representation.

Arabic metadata workflows are not edge cases. They represent everyday operational realities across academic, national, and public libraries in the region. As such, they provide a valuable lens through which to examine both the opportunities and the limits of AI in library systems, including the interplay between automated workflows and human oversight.

Arabic metadata as a multilingual workflow

Cataloging in many Middle Eastern libraries is inherently multilingual. Records are typically created in Arabic and English, and in some cases French. This creates a dual responsibility: maintaining consistency within each language while ensuring coherence across scripts.

Within OCLC cataloging environments such as Connexion client, this multilingualism is embedded in the bibliographic record structure. Arabic script fields are paired with romanized equivalents, following established transliteration standards such as ALA-LC. A single intellectual work may therefore exist in parallel representations that must remain aligned over time.

There are also technical considerations. Cataloging in Arabic depends on appropriate input methods, keyboard configurations, and support for non-Latin scripts within MARC environments. These infrastructural elements directly affect both efficiency and data quality.

Transliteration sits at the core of this workflow. While systems provide automated support, outputs frequently require manual correction. Arabic is highly context-sensitive, and small variations in spelling can significantly alter meaning.

Transliteration remains a strong candidate for improvement. Current approaches are largely rule-based. AI models, especially those trained on high-quality bilingual corpora, could offer more context-sensitive transliteration suggestions. However, these would still require validation, reinforcing the need for human oversight.

Discovery expectations and normalization practices

On the discovery side, user expectations introduce an additional layer of complexity. Arabic users expect search systems to handle orthographic variation seamlessly, without requiring precise input.

Recent enhancements in WorldCat Discovery illustrate how this is achieved through rule-based normalization. These include treating diacritics and non-diacritics as equivalent, normalizing character variants (such as different forms of alef), handling prefixes like “ال”, and ignoring elongation characters. Sorting rules are also adapted to reflect Arabic linguistic conventions.

What appears as simple search functionality is underpinned by carefully designed and tested normalization rules. These are deterministic and transparent, refined over time based on real usage patterns.

This is an important benchmark for AI. Any AI-driven approach to metadata creation or discovery must match or exceed this level of linguistic precision. Moreover, AI could extend normalization into cataloging workflows. While discovery systems normalize at query time, catalog records themselves often retain inconsistencies, particularly in legacy data. Machine learning models could assist in identifying and aligning variant forms across large datasets.

From local workflows to shared discovery infrastructures

A broader shift is also taking place in how Arabic collections are positioned within global discovery systems. Aggregated discovery initiatives, such as shared Arabic-language catalogs built on WorldCat infrastructure, reflect a move away from isolated local systems toward more integrated and visible ecosystems.

Field discussions with libraries across the Middle East point to a consistent concern: the global visibility of Arabic scholarship. Fragmentation in discovery and inconsistencies in metadata continue to limit access and representation.

From a strategic perspective, this aligns with broader discussions on the “collective collection” and the role of shared infrastructure in improving resource discovery. AI, when combined with such infrastructure, could help improve metadata consistency at scale and support cross-institutional alignment.

Risks and limitations

Despite these opportunities, the use of AI in Arabic metadata workflows raises several important concerns:

Language bias remains a significant issue. Many AI models are trained predominantly on English-language data, leading to uneven performance in Arabic. This reflects broader critiques of AI systems as reproducing existing linguistic and cultural imbalances.
Transliteration introduces additional risks. While rule-based systems are predictable, AI-driven approaches may produce variable outputs that are harder to standardize. This variability can undermine authority control and consistency.
There is also the risk of losing semantic nuance. Arabic names and terms often carry cultural and contextual meanings that may not be captured by automated systems.
Normalization itself must be approached carefully. Rule-based normalization is controlled and transparent. AI systems, by contrast, may over-normalize, removing distinctions that are meaningful within the data.

Shared implications for AI implementations and professional practice

Arabic metadata workflows reinforce several broader insights that emerged from OCLC Research’s earlier examination of AI and metadata management.

First, hybrid models are likely to be the most effective. AI can improve efficiency and scalability, but it does not replace the need for professional expertise. Human validation remains essential, particularly in linguistically complex environments. The earlier OCLC Research findings corroborate this, noting “the importance of designing AI implementations as enhancements to human expertise rather than replacements, ensuring that professional development pathways remain robust while leveraging AI’s potential to handle volume and routine tasks.”

Second, there is a need to move beyond English-centric assumptions in system design. Supporting multilingual knowledge infrastructures requires deeper engagement with linguistic diversity at the level of data, standards, and workflows. Again, this observation from Arabic metadata workflows finds a parallel with our earlier findings, which emphasize that “AI systems often lack the deep contextual understanding needed for community-specific terminology or cultural nuances that don’t appear in general training databases.”

Third, metadata enrichment is an area of growing interest. Many Arabic collections lack detailed subject metadata. AI could support the generation of subject headings, summaries, and keywords in Arabic. Our earlier findings also noted the opportunity AI affords for metadata enrichment: for example, institutional repository deposit processes “often fail to supply complete and accurate metadata because students and researchers find metadata creation burdensome and time-consuming.” AI-powered support in areas like subject heading suggestion or automated abstract generation can help close that gap.

Fourth, AI could contribute to backlog reduction by generating draft records or recommendations. This use case was also highlighted in OCLC Research’s earlier findings on AI and metadata workflows: “AI-generated brief records for these materials can enable them to appear in discovery systems earlier, accelerating the process of making hidden collections discoverable and supporting local inventory control. This approach addresses the immediate need for discovery while allowing records to be completed, enriched, or refined over time.”

Arabic metadata workflows present unique features that differ from English language-based systems, which in turn impact specific use cases for AI implementations. Yet as the preceding examples illustrate, there is also general perspective regarding AI-powered metadata workflows that applies equally to Arabic and non-Arabic systems alike. Perhaps most important is the observation that in considering AI implementations, the goal is augmentation rather than replacement, supporting catalogers in focusing their expertise where it adds the most value. There is a tendency in current debates to frame AI adoption as a binary choice between automation and professional control. But this framing is limiting. AI is more usefully understood as part of a continuum of human–machine collaboration, where the question is not just whether to use AI, but how, where, and under what constraints.

Looking ahead

A pragmatic approach is emerging across libraries in the Middle East. Institutions are exploring targeted AI applications, particularly in normalization, enrichment, and transliteration, while maintaining strong human oversight.

There is also an opportunity for collective action. Improving Arabic-language training datasets, strengthening authority control frameworks, and promoting collaboration across institutions will be critical for making AI effective in this space.

Developments in systems such as Connexion and WorldCat Discovery show that progress is already underway. AI can accelerate this work, but only if it is grounded in real workflows and informed by linguistic and cultural expertise.

Ultimately, this is not only a question of efficiency. It is a question of representation. Ensuring that Arabic knowledge is accurately described and fully visible within global discovery systems remains a central challenge and a meaningful test of how inclusive our infrastructures truly are.

The post No edge case: Understanding AI opportunities through Arabic metadata workflows appeared first on Hanging Together.

LibraryThing (Thingology): Author Interview: Katie Holt

2026-05-06T15:34:48+00:00

LibraryThing is pleased to sit down this month with author and romance enthusiast Katie Holt, whose 2024 debut, Not In My Book, follows the stormy relationship between a romance novelist and a literary author who are forced to work together on a book. A New York City resident and Tennessee native, Holt earned her degree in English and Creative Writing at NYU, while working at famed Manhattan bookstore, The Strand. When not writing, she works at St. Martin’s Press as an assistant editor. Her second novel, The Last Page, about the romance between a bookseller who dreams of running the bookstore where she works and the owner’s oblivious grandson who has inherited everything, is due out from Alcove Press later in May. Holt sat down with Abigail this month to discuss her new book.

How did the idea for The Last Page first come to you? Were you inspired by your own time as a bookseller? (Full disclosure: I also worked at The Strand for a number of years). Were other bookstore romances, like the one in You’ve Got Mail, an influence?

A fellow Strand-er!! That’s so cool! We’ll definitely have to see if we crossed paths with any of the same booksellers.

I was definitely inspired by my time there. I think it’d be impossible not to be! The Strand has a rich history and while I worked there, lots of the booksellers told me stories about Ben McFall. They had nothing but wonderful things to say about them and it was obvious he left such a deep mark on The Strand. I thought a lot about what it meant to leave behind a legacy in a bookstore. To be surrounded by books and for him to really be the most talked. I never met him, of course, so Leo was really inspired by my late grandfather. He was the coolest person I’ve ever met in my entire life and was a huge reader and supporter of literature.

I’m sure you know that the group of booksellers at The Strand were opinionated and eclectic and I loved it. I’m still friends with so many booksellers there and adore them. The bookseller relationships and dynamics were inspired by them for sure.

Nora Ephron is one of my greatest inspirations, so totally! You’ve Got Mail also has such a fun crew that are all different from each other, but they come together to create some hilarious scenes. I’ve always loved Notting Hill, too, and how that small shop just seems to be bursting with books. The Last Page has plenty of room in their store, but in my mind, books are just spilling over every edge of every surface.

You’ve been a romance advocate in both your professional and academic life, arguing to your college professors that the genre was worthy of consideration. What makes romance so special to you, and why should readers pay attention to it?

Oh goodness, I could wax poetic for days. I always say my favorite part of the romance genre is the HEA (happily ever after). That readers are guaranteed a safety net and it allows writers to explore some really heavy topics! The Last Page discusses grief throughout—what it means to different people, how it can ebb and flow, and how it doesn’t always make sense. Henry is also someone who suffers from depression and feels really embarrassed about it. I loved that I could discuss these topics the way I wanted to and tell the reader, “Don’t worry. They’re safe and you are, too.”

A lot of the time in college, I was told that these character arcs or plots I wrote were devalued by interweaving a love story that ends happily. I had one professor who always said, “I don’t buy it. I don’t buy that all the characters you write end up together.” Well, they do! It always blew my mind. Why would love decrease the value of anything? Isn’t it beautiful that you can evolve and change and grow and fall in love with someone who’s watched all of that happen? Who loves you before, after, and throughout?

There’s also something really special about how the romance genre varies so much within the genre. To me, it demonstrates how everyone views love differently and how everyone loves differently. I just think it’s beautiful and something special and worthy of celebrating.

The Last Page features a bookseller from New York City and a bookstore owner from Tennessee—a profession and two places that have been important to you personally. Are there specific spots or incidents in Ella and Henry’s tale that were inspired by your own life story?

I’m definitely inspired by my life in the city, and it’s impossible for me to not write about the things happening around me. (However, I unfortunately didn’t fall in love with a hot, shy, nerdy former football player while working at The Strand). New York is such a romantic city and I work hard to make the city feel like a character in the book. A reader
tagged me in a NOT IN MY BOOK book tour they did across NYC when they were visiting and I burst into tears. Those pictures of them at Peculiar Pub and Washington Square Park were exactly why I include real restaurants and locations. Specifically in this book, Ella and Henry go to Kingston Hall, which is one of my favorite places. It’s buy one get one free beer on the weekends!! And there’s a great pool table and such interesting interior design. I’ve brought my laptop there plenty of times to get some writing done and thought it was such a cool place for them to stumble into.

Although I loved being a bookseller, I did have the quintessential yet frustrating “I saw a book with a black cover here three weeks ago. Can you help me find it?” which I include in The Last Page. Bill Clinton also did come in and sift through American History. There wasn’t a naked man that ran through the store—he had on a speedo in real life.

Something that was really inspired by my life was Leo. Like I said, he was based on my grandfather who I was really close to. He had been pretty sick for a few years and every time I saw him, I braced myself because it could’ve been the last. The thought of losing him haunted me and I think I wrote this in a place of nearly preemptive grief? I kind of think of it like a letter from my past self telling me that I’ll survive losing him, even if it hurts like hell sometimes. He never got to hold a copy of The Last Page, though, I know he would’ve been proud and read every single word (as he mortifyingly told me he did with Not In My Book).

Tell us a little bit about your writing process—how and where you like to write, and how you construct your stories. Are you a plotter or a pantser?

I have to plot. I recently finished writing a book that I tried to pants and it was horrible. I hated every single second of it, deleted everything, and started over.

My process has evolved as my life has! For Not In My Book, I was home for the pandemic and wrote the first draft in my childhood bed and my parent’s kitchen table. I had to have my Beats on and a bag of BBQ chips and Kombucha by my side (in my head they counteracted each other??).

Nowadays, though, I love to write in cafes or bars in the city. I’ve always thought it was very Cool Girl when I saw people writing or reading at bars. Recently, I’ve been grabbing my notebook and physically writing out scenes or chapters. I also love to listen to music while I write (mostly Taylor Swift) and think about how the characters fit that song or if they ever have. Because Rosie definitely listened to “Welcome to New York” for a month straight when she moved to New York and Ella definitely had “Honey” on a constant loop.

What advice would you give to up-and-coming authors, particularly up-and-coming romance authors?

When I was in high school, I took a writing course with Rachel Carter and it completely transformed the way I wrote and my perspective on it. She told me to write as much as I read and read a lot. The only way to know the mechanics of the genre or even a book is to completely entrench yourself in it. Read lots and lots of books and write every day. Even if it’s just a sentence! Writing is a muscle and if you don’t exercise it, it’ll weaken.

I also think it’s important to give yourself grace. No one is publishing their very first draft. It’s okay for your writing to be terrible and make sense to no one else but you. Sometimes it’s all in the revisions! But don’t be too hard on yourself.

What’s next for you? Do you have any books currently in the works?

I have a couple of things in the works!! I’m working on two books right now that are very different from each other. I don’t want to reveal too much just yet, but these really feel like the stories I’ve been waiting to tell for sometime, so I’m super excited about them!

Tell us about your library. What’s on your own shelves?

All kinds of romance! I really don’t shy away from any of the subgenres. I’ve been into speculative and dark recently, but I’m hoping historical makes a resurgence, especially now that Lisa Kleypas is back!

It’s really rare I read outside of romance, but that’s my small goal for this year. I’ve always been a physical reader, but I’m dabbling in some book club fiction and nonfiction through my audiobooks.

What have you been reading lately, and what would you recommend to other readers?

I just finished Tessa Bailey’s new time travel romance, Broken Rival, and I feel like I’m having a parasocial relationship with them. Like…it’s that serious. A Little Buzzed by Alys Murray was so intensely sexy that my Apple Watch asked if I was working out because my heartbeat was so high. That is a MUST read! I also read The Heartbreak Hotel by Ellen O’Clover recently and was stunned by it and the musings on love and life after a breakup. The characters are so well drawn and she strikes the exact right balance of emotionally intense and absolute yearning. Secret Nights and Northern Lights by Megan Oliver is also what second chance romance dreams are made of. I cannot wait to read more from her.

David Rosenthal: The Permissionless Catch-22

2026-05-05T15:46:01+00:00

Potential Attack Target

Suppose some genre of content is under attack by powerful adversaries. Lets take political satire as a thought experiment in which powerful politicians are attacking sites and Web archives hosting it by sending bogus DMCA takedowns, suing for defamation, buying up their hosting platforms, getting their flying monkeys to flood them with spam, and so on. Below the fold I discuss the problem facing the defense.

White Hats

You and a few friends get together to fight back. You think about setting up the not-for-profit LOOPS (Library Of Offensive Political Satire) to collect and preserve it, but quickly realize it would be immediately sued into bankruptcy.

Inspired by BitTorrent, the alternative you see is to write and distribute the software for a permissionless, peer-to-peer network. It would use erasure coding to ensure that each node held only a fraction of any individual satire, but each satire was held in aggregate by many nodes. That promises no central point subject to legal attack, and no node holding an identifiable part of a satire.

The graph of the security of such a system against the number of nodes will be an S-curve. With a small number of nodes it isn't secure. As the number increases, security increases slowly until a point where each of the average satire's chunks are held on at least two nodes. Then it rises rapidly until the law of diminishing returns kicks in and the curve flattens out.

Thus you need many nodes under independent control. You need to motivate many people or institutions to run a node. The lower the capital (capex) and operational (opex) costs of doing so, the more likely you are to get to the steep part of the curve in time to foil the looming attacks.

Catch-22

The security of a permissionless peer-to-peer system depends upon the cost of mounting an attack being greater than the reward for a successful attack. But making it cheaper and easier to run a node has the inescapable unintended consequence of making it cheaper and easier to mount an attack. Catch-22!

Black Hats

The white hats have to pay the capex then pay the recurring opex indefinitely. In most cases the black hats can mount a one-time attack. Because it is cheap and easy to spin up a node, the black hats can avoid paying the capex and pay the opex only one time by renting a large group of nodes from AWS.

But the black hats don't even have to do this. The white hats have built a permissionless system that, like almost all such systems, is not actually decentralized. The black hats can mount a software supply chain attack and, instead of creating new nodes, compromise the white hats' existing nodes.

This works because, as we see in almost all P2P systems, no-one is willing to pay the cost or devote the unpaid effort to doing clean-room re-implementations of the software. Especially since making nodes secure but cheap and easy to run is a difficult engineering problem. Even if there are multiple independent implementations, there are two further problems. First, it is likely that they all share common dependencies on libraries that could be targets for the attack. Second, network effects mean that one of the implementations will capture the bulk of the market.

Conclusion

The need to make running a node cheap and easy isn't just a Catch-22, it is a Catch-22 that favors the black hats.

Open Knowledge Foundation: Brazil and OKFN partner to enable citizens to ‘talk’ to national data

2026-05-04T17:03:56+00:00

Together, we are developing a tool that will bring public data closer to citizens, allowing them to get data by simply asking a question.

The post Brazil and OKFN partner to enable citizens to ‘talk’ to national data first appeared on Open Knowledge Blog.

Library | Ruth Kitchin Tillman: OpenRefine: API Basics and Repeat Operations

2026-05-04T00:00:00+00:00

OpenRefine has an API. Not just for reconciliation, but an API you can use to perform actions from creating/deleting a project to blanking down columns and mass editing. Even some of the more experienced users I’ve asked were surprised by this.

I think if you’d asked me before this month, I’d have said “Maybe?” I’ve spent enough time in the documentation to see it mentioned, but hadn’t spent any time investigating.

Like many OpenRefine users, much of what I do involves looking at the data, considering what I need to do next, performing an action or two, evaluating the outcome, repeat. That doesn’t lend itself well to API work. But sometimes, as I’ve mentioned in the previous blog posts, I’ve been doing a lot of repeating tasks at the start of projects. Once those are performed, I need to visit the project sheet and assess the data. So when I was first trying to figure out how to repeat operations, I decided to learn more about the API.

API Basics

Because OpenRefine runs on a local server (or you could host an instance on an actual server), you can send a pretty straightforward set of GET or POST actions at that URL. You can even create projects using the API, with appropriate parameters, though I haven’t yet teased that out. Data has to be sent as multipart/form-data.

While I’m going to be using the “perform operations” function, which is documented, I noticed a lot of actions in the CSRF writeup that were missing in the official API documentation. When I have some free time – if I have some free time – I may try these out.

Authentication

Even if you’re running it locally, you’ll need to get a CSRF token for POST requests which change the data in anyway.

To handle this in my fuller Python script, I wrote/repurposed the following code:

def auth_me():
  '''auth function broken out'''
  response = requests.get(server + "/command/core/get-csrf-token", params={"project":project_id})
  access = response.json()["token"]
  return access

Performing Operations With the API

Performing operations is a simple POST. This is my very simple Python, which only requires requests and json libraries (there are actual OpenRefine clients, but the Python client’s GitHub page was archived by its owner so I didn’t want to rely on it). The code is commented, but essentially I:

Set the server
Set the project ID
Define and perform the auth
Paste in the operations copied as described in the previous post and dumps it into JSON. Critically, however, I had to change false to False or Python got mad. I don’t know if it would be better to put “false” in quotes, that’s another thing to test more (maybe using a case which should be True, or where false is more noticeable). I didn’t need to repeat and this works.
Using requests, post to perform-operations url. Pass on parameters of the project ID, the csrf key, and the actual operations I want to perform. For the sake of code lenght, this is a much shorter set of operations than I was actually performing.
Process the result and print something which helps me understand if it was successful or not.

import requests, json

## set variable on the off-chance it's a different server sometime
server = "http://127.0.0.1:3333"

## could be set as an argument. could also be redone as a list of IDs and iterate through them to perform the same functions (adjustments needed below)
project_id = "18117329352"

## gets new auth code each time, I don't know when they expire
def auth_me():
  '''auth function broken out'''
  response = requests.get(server + "/command/core/get-csrf-token", params={"project":project_id})
  access = response.json()["token"]
  return access

## now do actual auth
key = auth_me()

## simple pasted in entire output and changed "false" to "False" and then threw into a json.dumps
post_operations=json.dumps([
  {
    "op": "core/blank-down",
    "engineConfig": {
      "facets": [],
      "mode": "row-based"
    },
    "columnName": "Catalog Key",
    "description": "Blank down cells in column Catalog Key"
  },
  {
    "op": "core/text-transform",
    "engineConfig": {
      "facets": [],
      "mode": "row-based"
    },
    "columnName": "Title",
    "expression": "grel:value + \" - \" + row.record.index",
    "onError": "keep-original",
    "repeat": False,
    "repeatCount": 10,
    "description": "Text transform on cells in column Title using expression grel:value + \" - \" + row.record.index"
  }
])

## now perform the operation
perform_operations = requests.post(server + "/command/core/apply-operations",
  params=
    {
      "project":project_id,
      "operations":post_operations,
      "csrf_token":key
      })

## oh my god tell me what happened
if perform_operations.status_code == 200:
  if perform_operations.json()["code"] == "ok":
    print("Operations performed successfully")
  else:
    print(json.loads(perform_operations.json()))
else:
  print("Status code:",perform_operations.status_code)

Evaluating Efficiency

As a one-off, I don’t think using the API for this purpose is going to be more efficient than simply downloading a copy of the operations I want to replicate and uploading the file to each new project.

What would tip it over into efficiency for me would be if I could figure out how to create projects, do those as a batch, get the IDs, and then run a second step to perform all the operations on them as a list.

As I said at the beginning, I think a major reason everything from repeating functions to how to use the API isn’t more widely-known among regular practitioners like myself is that we’re so rarely doing something with this kind of repeated process. I don’t think I’ll find the API helpful for most things I do. But when I’m on a big project like this one? I’ll keep exploring.

Ed Summers: Weekly Bookmarks

2026-05-03T04:00:00+00:00

These are some things I’ve wandered across on the web this week.

🔖 Soundcamp / Reveil

2026 is the 40th year since the ecology park was set up on abandoned dockland. A free programme of walks, talks and workshops will be a chance to find out more about histories of the site and join a conversation about its future. Listen in the PITCH soundtent as the Reveil 24+1hr broadcast loops the earth at daybreak. Join local naturalists for a bat walk on Saturday night and a bird walk at dawn on Sunday.

🔖 endetymes

the Ende Tymes Festival brings you the best in noise and experimental music since 2011

🔖 Gray Area

Our mission is to cultivate, sustain, and apply antidisciplinary collaboration — integrating art, technology, science, and the humanities — towards a more equitable and regenerative future. Since our inception in 2008, Gray Area has established itself as a singular hub for critically engaging with technology and culture in the Bay Area, while also reaching a global audience. Through our platform of public events, education, and research programs we empower a diverse community of creative practitioners with the agency to create meaningful social impact through category-defying work.

🔖 RubyLLM

One beautiful Ruby API for GPT, Claude, Gemini, and more. Easily build chatbots, AI agents, RAG applications, and content generators.

🔖 Contributor Poker and Zig’s AI Ban

Unfortunately the reality of LLM-based contributions has been mostly negative for us, from an increase in background noise due to worthless drive-by PRs full of hallucinations (that wouldn’t even compile, let alone pass CI), to insane 10 thousand line long first time PRs. In-between we also received plenty of PRs that looked fine on the surface, some of which explicitly claimed to not have made use of LLMs, but where follow-up discussions immediately made it clear that the author was sneakily consulting an LLM and regurgitating its mistake-filled replies to us.

To be clear, the point here is not to say that we believe that this is all that AI is. We don’t. This is clearly a misuse of the tool, but it is also what the overwhelming majority of LLM-based contributions looked like for our project.

So while one could in theory be a valid contributor that makes use of LLMs, from the perspective of contributor poker it’s simply irrational for us to bet on LLM users while there’s a huge pool of other contributors that don’t present this risk factor.

🔖 We did the math on AI’s energy footprint. Here’s the story you haven’t heard.

AI’s integration into our lives is the most significant shift in online life in more than a decade. Hundreds of millions of people now regularly turn to chatbots for help with homework, research, coding, or to create images and videos. But what’s powering all of that?

Today, new analysis by MIT Technology Review provides an unprecedented and comprehensive look at how much energy the AI industry uses—down to a single query—to trace where its carbon footprint stands now, and where it’s headed, as AI barrels towards billions of daily users.

🔖 Before GitHub

As much as I like the idea of things fading out of existence, we absolutely need libraries and archives.

Regardless of whether GitHub is here to stay or projects find new homes, what I would like to see is some public, boring, well-funded archive for Open Source software. Something with the power of an endowment or public funding to keep it afloat. Something whose job is not to win the developer productivity market but just to make sure that the most important things we create do not disappear.

The bells and whistles can be someone else’s problem, but source archives, release artifacts, metadata, and enough project context to understand what happened should be preserved somewhere that is not tied to the business model or leadership mood of a single company.

🔖 Do I belong in tech anymore? On quitting, the spread of AI, and the loss of an ideal.

Ironically, what I’ve gained from AI is a deeper appreciation for human communication, in all its messy imperfection. The point of a code review is not simply for good code to make it into a codebase, but to build institutional knowledge as people debate and iterate and compromise, slow as it may be. Friction is good.

🔖 Getting my daily news from a dot matrix printer

For a while now I’ve started my day by unlocking my phone and scrolling through different news and social media sites to see what’s going on. It’s not exactly great for my mental health and I’ve been trying to cut down on screen time for a while. I still want to stay up-to-date though, especially after I get up in the morning.

I recently purchased a dot matrix printer from eBay, and thought it would be a great excuse to have a custom “front page” printed out and ready for me each day. So, that’s what I built!

🔖 Claude-powered AI coding agent deletes entire company database in 9 seconds — backups zapped, after Cursor tool powered by Anthropic’s Claude goes rogue

The founder of PocketOS has penned a social media post to warn others about the “systemic failures” of flagship AI and digital services providers. Jer Crane was inspired to write a public response after an AI coding agent deleted his firm’s entire production database. The AI agent’s misdemeanors were then hugely amplified by a cloud infrastructure provider’s API wiping all backups after the main database was zapped. This tag team of digital trouble has wiped out months of consumer data essential to the firm’s, and its customers, businesses.

🔖 WARC Extensions: From idea to stable specification

A brief presentation from Alex Osbourne about how to manage up dates to the WARC ISO standard.

🔖 Twenty Years Later Twitter, Transformation, and Platform Legacies

Several tensions run through this issue, then, starting with the pull between nostalgia and critique. Resisting the temptation to romanticise old Twitter requires acknowledging what was genuinely lost without pretending the platform was ever perfect. Twitter was always a bit like Schrödinger’s cat of communication, containing both possibility and harm, connection and toxicity, democratic potential and algorithmic manipulation. The transformation under Musk has made some of these dynamics far more visible, but it is important to acknowledge that they were present from the beginning.

The question of what to call the platform carries political and analytical weight. Continuing to say “Twitter” rather than “X” matters. “Twitter” encapsulates a set of communication practices, a vernacular, a mindset, and a scholarly object that cannot be erased by new ownership. The archive remains, at least in part, the research corpus persists, and the cultural imprint endures even as the platform morphs into a very different beast.

🔖 Speaker notes from Resisting Big Tech Empires

Decomputing is about deautomatisation; about extracting ourselves from the patterns of machinic relations which are amplified by tech. It’s about mutual aid that comes from the recognition of mutual vulnerability, and care that doesn’t depend on classifying people according to algorithmic boundaries. Decomputing adopts the approach to tech development outlined in Illich’s tools for conviviality; developing tools that enable autonomy and adaptation, rather than the conditioned responses demanded by manipulative systems.

🔖 Safety-I and Safety-II: The Past and Future of Safety Management

Focusing on what goes right, rather than on what goes wrong, changes the definition of safety from ’avoiding that something goes wrong’ to ’ensuring that everything goes right’. More precisely, Safety-II is the ability to succeed under varying conditions, so that the number of intended and acceptable outcomes is as high as possible. From a Safety-II perspective, the purpose of safety management is to ensure that as much as possible goes right, in the sense that everyday work achieves its objectives.

🔖 Non serviam

Non serviam is Latin for “I will not serve”. The phrase is traditionally attributed to Satan, who is thought to have spoken these words as a refusal to serve God in Heaven. Today, it is used as a motto by a number of political, cultural, and religious groups to express their wish to rebel, or simply not serve. It may be used to express a radical view against established beliefs and organizational structures accepted as the status quo.

🔖 What are embeddings?

Peter Norvig urges us to teach ourselves programming in ten years. In this spirit, after several years of working with embeddings, foundational data structures in deep learning models, I realized it’s not trivial to have a good conceptual model of them. Moreover, when I did want to learn more, there was no good, general text I could refer to as a starting point. Everything was either too deep and academic or too shallow and content from vendors in the space selling their solution. So I started a project to understand the fundamental building blocks of machine learning and natural language processing, particularly as they relate to recommendation systems today. The results of this project are the PDF on this site, which is aimed at a generalist audience and not trying to sell you anything except the idea that vectors are cool. I’ve also been working on Viberary to implement these ideas in practice

🔖 Talk Talk - Live at Montreux 1986

From Aquarium Drunkard:

By July 1986, Talk Talk were still a functioning live unit touring behind The Colour of Spring. But something had already shifted as evidenced by this set from that summer’s Montreux Jazz Festival. Listen closely and you can hear the architecture beginning to loosen: tempos breathe, arrangements open, and familiar material begins to drift toward something less fixed, less performative.

This would be their final tour. Within a year, Mark Hollis and company would retreat into the studio to begin work on Spirit of Eden, a record that all but rejects the idea of live translation. As such, this Montreux performance exists as a kind of threshold document, one that captures the band onstage one last time before the music folds inward on itself.

🔖 The West Forgot How to Make Things. Now It’s Forgetting How to Code

The skills you need to be effective now are different. Technical expertise alone isn’t enough anymore. You need people who can take ownership, communicate tradeoffs, push back on bad suggestions from a machine that sounds very confident. Leadership qualities. Our last hiring round tells you how rare that is: 2,253 candidates, 2,069 disqualified, 4 hired. A 0.18% conversion rate. The combination of technical skill and the judgment to know when the AI is wrong barely exists in the market anymore.

🔖 There Will Be a Scientific Theory of Deep Learning

In this paper, we make the case that a scientific theory of deep learning is emerging. By this we mean a theory which characterizes important properties and statistics of the training process, hidden representations, final weights, and performance of neural networks. We pull together major strands of ongoing research in deep learning theory and identify five growing bodies of work that point toward such a theory: (a) solvable idealized settings that provide intuition for learning dynamics in realistic systems; (b) tractable limits that reveal insights into fundamental learning phenomena; (c) simple mathematical laws that capture important macroscopic observables; (d) theories of hyperparameters that disentangle them from the rest of the training process, leaving simpler systems behind; and (e) universal behaviors shared across systems and settings which clarify which phenomena call for explanation.

Web Science and Digital Libraries (WS-DL) Group at Old Dominion University: 2026-05-02: Evolution of Language-Guided Control in UAV Systems

2026-05-02T22:20:15+00:00

Evolution of Language-Guided Control in UAV Systems

From Search and Rescue (SAR) to intelligence, surveillance and reconnaissance (ISR), the use of small unmanned aerial systems (sUAS) in scientific and military applications has become increasingly more prominent. One of the most important aspects of the use of these devices is how they are directed. This blog takes a look at the evolution of control for these systems from 2017 to 2026, by examining the approach used in the paper, “'Fly Like This': Natural Language Interface for UAV Mission Planning” [1], to the use of the Large Language Models (LLMs) being examined in “A Universal Large Language Model-Drone Command and Control Interface” [2]. These papers explore how natural language interfaces (NLI), or more generally, each of these multiple human-system interfaces can be used to dictate the actions of these platforms. Taken together, these publications reflect the technological capabilities present at the time of their release, illustrating nearly a decade of progress. Evaluating the time span between these papers, one must consider the future trajectory and ask the question, what's next?

Fly Like This (2017)

In this paper, published in March of 2017, the authors perform a Human-System Integration (HSI) evaluation of the effectiveness, efficiency and natural comfort of each of the multiple interfaces, such as, speech, hand gesture, and traditional mouse, as input to define the flight path of an Unmanned Aerial Vehicle (UAV). Each interface had the objective of identifying one of 12 flight path segments expressed by the user. As each segment is generated by its respective interface, they are linked together to form the UAV’s planned flight path.

Figure 1. Gesture library of 12 trajectory segments developed by Chandarana et al. [3]

Interfaces

Gestures

Using a Leap Motion Controller (now known as Ultraleap), and accompanying SDK v2.2.6, 12 hand gestures, using only their right hand, were converted into 12 trajectory segments to produce the UAV flight path. The controller consists of three infrared cameras to produce sub-millimeter accuracy in an 8 square-foot interactive volume. The hand motions were chosen to reflect the overall motion the user expected the UAV to take. For example, if the user intends for the UAV to form an orbit, they create a circular pattern with their hand. Each segment was confirmed with either a right for yes or a left for no.

Figure 2: Leap Motion Controller (available in 2017)

Speech

The development of the speech interface used a commercial-off-the-shelf (COTS) microphone and processed that data using Carnegie Mellon University’s (CMU) Sphinx software with its built-in US-English acoustic and language models. Sphinx is a speech recognition software designed to take digitized acoustic signals and convert them to text. Software was developed to convert words such as “right”, “left”, “circle”, and “spiral” to UAV flight path segments. Each segment was confirmed with a verbal “yes” or “no”.

Figure 3: Audio-Technica PRO 8HEmW

Mouse

As a baseline, a mouse interface was used in conjunction with a drop-down menu to generate the 12 trajectory segments. Using a traditional computer mouse, the user would select the type of segment they wanted to be generated, followed by a pop-up window to confirm with a yes/no selection to add that segment to the flight path.

Accuracy and Speed

Of the three interfaces mentioned above, how fast can human intent be translated into actionable parameters to develop a flight path? The paper shows that in both accuracy and speed, the mouse reigned supreme by a relatively large margin.

Figure 4: Comparison of mouse, speech, and gesture input for UAV flight path entry from “Fly Like This: Natural Language Interfaces for UAV Mission Planning” [1]

In a real-world operational environment, the operator faces a significantly higher cognitive load than that represented in a controlled study. Beyond the mechanical act of inputting coordinates or gestures, the operator must first perform situational assessment and path optimization. This "pre-input" phase represents a substantial cognitive overhead that is often decoupled from the efficiency of the interface itself.

The Pre-LLM Conclusion

The study described above is a representative sample of the state of the art in 2017 and at that time, the underlying technology behind tracking speech and hand gestures became advanced enough to generate quality data to produce UAV pathing with relatively few errors. Even today, the implementation of such systems still remains technically impressive. However, fast forward to late 2022 and the emergence of publicly available LLMs hit the stage. This development substantially advanced human system integration, specifically with respect to the capabilities and applications of natural language processing technologies.

A Universal Large Language Model-Drone

Command and Control Interface (2026)

ChatGPT launched publicly in November of 2022, marking the moment Large Language Models became mainstream and publicly available. These LLMs began to approximate human intent by pattern matching text input and generating aligned responses. LLMs not only process human language, but also analyze a seemingly inexhaustible list of data structures, including JSON.

In 2025, the Model Context Protocol (MCP) was introduced and subsequently adopted by numerous LLM platforms, enabling these models to access a range of tools that enhance their functional capabilities. In the paper “A Universal Large Language Model – Drone Command and Control Interface” [2], the authors describe a methodology to enable LLMs, through an MCP Server, to control drone behavior. This represents a significant step forward in translating human intent to UAV action or behavior.

The Architecture & Interface

Figure 5: System architecture of the MCP-based drone control interface from
“A Universal Large Language Model -- Drone Command and Control Interface” [2]

The authors demonstrated this concept by constructing their architecture using three primary components, an LLM, and MCP Server, and a drone, both virtual and physical. The LLM at this point has become an interchangeable component and therefore multiple LLM providers such as OpenAI, Anthropic, and Gemini are all capable of performing in this architecture. These LLMs aren’t restricted to connecting to a single MCP server, making for a modular system that allows the system architect to pick and choose the best MCP servers for the system they’re constructing. The authors connected the Google Map MCP server to the LLM to provide the model with navigational information, while the ‘drone controlling’ MCP Server was a custom server called ‘droneserver’, which is freely provided along with the source code on GitHub. Combined, these two servers provided the LLM with the information and control required to produce Micro Air Vehicle Link (MAVLink) messages. MAVLink is a common standard used in the drone community which provides command and control, as well as telemetry from the drone. The third component of this architecture can either be virtual or physical. For both the physical and virtual components of the study, the authors employed the control and simulation software ArduPilot. Initially the authors connected these MCP servers to a virtual drone controlled by ArduPilot in a virtual environment such as Gazebo. When the final component was used as a physical system, a drone equipped with a Raspberry Pi Zero W (with connection to the internet) had the MCP server installed on it locally and was provided an interface to the flight controller.

The Evolution of Objective-Based Control

A decade of progress has replaced rigid controls with total mission flexibility. Users have the option to zoom out to issue broad, high-level objectives or dive deep into the weeds of specific flight paths, choosing the exact level of abstraction the mission requires. For example, in an ISR application, a user can tell the LLM to search a given area and it can refer to the appropriate algorithm to maximize a specific goal such as fuel efficiency or coverage. Alternatively, a user could specify tight flight paths for a UAV to follow or even mix and match.

With the addition of the MCP server, the LLM gains access to additional information the user isn't required to specify or even consider. If the objective the system needs to maximize is fuel efficiency, the LLM could autonomously access the latest weather patterns allowing the drone to map a course over various altitudes capitalizing on tailwinds to increase the range and time of flight for the platform.

Ultimately, with the introduction of the LLM, the shift in control has gone from manual navigation to strategic oversight. By abstracting the technical burdens of UAV flight, these types of systems empower the operator to focus on the mission. Advancing forward, a single operator is no longer restricted to one or a few drones, but an entire swarm of drones who handle the user's intent.

- John Deasy

References

Saephan, M., Meszaros, E., Trujillo, A., & Allen, B. (2017). Fly like this: Natural language interfaces for UAV mission planning. In Proceedings of the Tenth International Conference on Advances in Computer-Human Interactions (ACHI 2017) (pp. 40–46). IARIA.

[2] Ramos-Silva, J. N., & Burke, P. J. (2026). A universal large language model -- drone command and control interface. arXiv. https://doi.org/10.48550/arXiv.2601.15486

[3] Chandarana, M., Meszaros, E. L., Trujillo, A. C., & Allen, B. D. (2016). Natural language and gesture-based interfaces for UAV mission planning. AIAA AVIATION Forum, Washington, D.C. https://ntrs.nasa.gov/citations/20160010163

LibraryThing (Thingology): May 2026 Early Reviewers Batch Is Live!

2026-05-01T18:28:38+00:00

Win free books from the May 2026 batch of Early Reviewer titles! We’ve got 247 books this month, and a grand total of 2,956 copies to give out. Which books are you hoping to snag this month? Come tell us on Talk.

If you haven’t already, sign up for Early Reviewers. If you’ve already signed up, please check your mailing/email address and make sure they’re correct.

» Request books here!

The deadline to request a copy is Tuesday, May 26th at 6PM EDT.

Eligibility: Publishers do things country-by-country. This month we have publishers who can send books to the US, the UK, Canada, Ireland, Germany, Italy, Australia, Spain, Poland, Sweden and more. Make sure to check the message on each book to see if it can be sent to your country.

Thanks to all the publishers participating this month!

Alcove Press	American Taboo Press	Autumn House Press
Bellevue Literary Press	City Owl Press	CMU Press
Crooked Lane Books	Cynren Press	Entrada Publishing
eSpec Books	Espresso Publishing House	Flat Sole Studio
Galaxy Press	Gefen Publishing House	Hawthorn Quill Publishing
Heritage Books	Hybrid Sequence Media	Identity Publications
Inferno Books	Infinite Books	It’s Alive! Books
Kinkajou Press	LaPuerta Books and Media	NeoParadoxa
OC Publishing	Picket Fire	Pocketbook Press
Prolific Pulse Press LLC	PublishNation	RIZE Press
Ronsdale Press	Rootstock Publishing	Running Wild Press, LLC
Shadow Dragon Press	Shilka Publishing	Silent Clamor Press
Simon & Schuster	Tundra Books	Type Eighteen Books
University of Nevada Press	University of New Mexico Press	Vibrant Publishers
W4 Publishing, LLC	What on Earth!

Digital Library Federation: DLF Digest: May 2026

2026-05-01T12:00:29+00:00

A monthly round-up of news, upcoming working group meetings and events, and CLIR program updates from the Digital Library Federation. See all past Digests here.

Hello DLF Community!

There’s growing momentum as the Call for Proposals for the 2026 Virtual DLF Forum officially opens, inviting contributions that reflect this year’s focus on practical strategies, community-grounded work, and shared challenges across the field. With recommendations from members of the Forum Program Committee and DLF Committee for Equity and Inclusion (CEI), which continues to advance inclusive and equitable practices across the GLAM community through open monthly meetings and a new Zotero resource library, we have expanded this year’s CFP to more intentionally include and center participation from Historically Black Colleges and Universities (HBCUs), Tribal Colleges and Universities (TCUs), Hispanic-Serving Institutions (HSIs), and Minority-Serving Institutions (MSIs).

We’re also introducing a new digital storytelling format that encourages partnerships between librarians, archivists, and community collaborators to share not just project outcomes but the relationships and processes behind the work, and to help attendees imagine how these approaches can be adapted in their own contexts (learn more about the new format on the DLF blog). We hope you’ll consider submitting a proposal and sharing the CFP across your networks!

Warmly,

-Shaneé

This month’s news

The CFP for the 2026 Virtual DLF Forum is Now Open: Submit your proposal by May 11 at 11:59 PM ET and help shape this year’s conversations.
DLF DAWG Accessibility Audit Training: Register now for a practical overview of the DLF DAWG Accessibility Auditing Shortlist on May 12, at 11 AM ET. D. Krahmer (DLF DAWG IT/Development Group co-chair) will explain a basic process for assessing software accessibility and how you can apply it at your own institution. Advance registration required; the session will be recorded.
RLUK/DLF joint event Copyright and Licensing in the Age of AI: Register and RSVP for an interactive transatlantic event hosted by CLIR+DLF’s Data and Digital Scholarship Working Group and Research Libraries UK’s Digital Scholarship Network. The session will explore how libraries can support and lead on copyright and licensing in the age of AI, taking place on May 14 at 11 ET.
Myth-Busting Guidance: ADA Title II Urban Legends: Sorting Fact from Fiction About the 2024 Updates, a blog authored by members of DLF’s Digital Accessibility Working Group, unpacks common misconceptions about ADA Title II compliance for the April 2026 deadline (Note: compliance deadlines have been extended through 2027 & 2028). Learn more in the full post on the DLF blog.
DLF Webinar Recording Available: The Content Authenticity and Provenance in the Age of Artificial Intelligence webinar recording and slide decks are now available on the blog for those who would like to revisit or catch up on the conversation. Visit the DLF blog for the replay.
Sustaining Rights Infrastructure: Digital Scholar has been announced as the new institutional home for RightsStatements.org. Read more about the change.

This month’s open DLF group meetings:

For the most up-to-date schedule of DLF group meetings and events (plus conferences and more), bookmark the DLF Community Calendar. Meeting dates are subject to change. Can’t find the meeting call-in information? Email us at info@diglib.org. Reminder: Team DLF working days are Monday through Thursday.

DLF Born-Digital Access Working Group (BDAWG): Tuesday, 5/5, 2pm ET / 11am PT.
DLF Digital Accessibility Working Group (DAWG): Tuesday, 5/5, 2pm ET / 11am PT USA.
DLF AIG Cultural Assessment Working Group: Monday, 5/11, 1pm ET / 10am PT.
AIG User Experience Working Group: Friday, 5/15, 11am ET / 8am PT
AIG Metadata Assessment Group: Friday, 5/15, 2pm ET / 11am PT.
DLF Climate Justice Working Group: Tuesday, 5/26, 3pm ET / 12pm PT.
DLF Open Source Capacity Resources Group: Wednesday, 5/27, 1pm ET / 10am PT.
DAWG Policy & Workflows: Friday, 5/29, 1pm ET / 10am PT.

DLF groups are open to ALL, regardless of whether or not you’re affiliated with a DLF member organization. Learn more about our working groups on our website. Interested in scheduling an upcoming working group call or reviving a past group? Check out the DLF Organizer’s Toolkit. As always, feel free to get in touch at info@diglib.org.

Get Involved / Connect with Us

Below are some ways to stay connected with the digital library community and us:

Subscribe to the DLF Forum newsletter.
Join, start, or revive a working group and browse their work on the DLF Wiki.
Subscribe to our community listserv, DLF-Announce.
Bookmark our Community Calendar.
Learn more about becoming a DLF member organization.
Follow us on LinkedIn and YouTube.

Contact us at info@diglib.org.

The post DLF Digest: May 2026 appeared first on DLF.

Artefacto: Five for Friday – AI policy examples for libraries

2026-05-01T07:11:09+00:00

The dust is settling, the bubble has yet to burst and more libraries than ever have their AI policy in place, for users, for staff or hopefully for both. If you’re still working out where to start, we hope this post can help. Many of these exist at an institutional level, where, for example, a [...]

Source

Dan Cohen: A Framework for Books and AI in the Public Interest

2026-04-30T20:52:59+00:00

“Color of Reading” by MarLeah Cole, CC BY 2.0

Two years ago, Dave Hansen, the Executive Director of Authors Alliance, and I wrote “Books Are Big AI's Achilles' Heel,” a piece on how the leading AI companies may have unimaginable sums of money and vast data centers, but are badly in need of what humble libraries have in abundance: books. Those companies, of course, understood this weakness and were trying to fill in the gap in any way they could. There are now dozens of lawsuits by authors and publishers against these tech firms for downloading and storing digitized books from the sketchier corners of the internet.

Dave and I proposed an alternative pathway, spearheaded by libraries and oriented not toward commercial uses but toward the public good:

A library-led training data set of books would diversify and strengthen the development of AI. Digitized research libraries are more than large enough, and of substantially higher quality, to offer a compelling alternative to existing scattershot data sets. These institutions and initiatives have already worked through many of the most challenging copyright issues, at least for how fair use applies to nonprofit research uses such as computational analysis. Whether fair use also applies to commercial AI, or models built from iffy sources like Books3, remains to be seen.
Library-held digital texts come from lawfully acquired books — an investment of billions of dollars, it should be noted, just like those big data centers — and libraries are innately respectful of the interests of authors and rightsholders by accounting for concerns about consent, credit, and compensation. Furthermore, they have a public-interest disposition that can take into account the particular social and ethical challenges of AI development.

Thanks to the Mellon Foundation, this planning project was funded, and we held workshops across the United States with librarians, scholars, technologists, authors, and publishers to imagine what such an initiative might look like, how it might function, and what it would take to bring it into existence. We’re delighted to release the final report from that yearlong study, The Public Interest Corpus: A Framework for Implementation, co-authored by Dave, Thomas Padilla, Giulia Taurino, and myself.

From the introduction:

The rapid advancement of artificial intelligence represents one of the most significant technological transformations of the twenty-first century, with profound implications for research, education, creativity, and civic life. Yet the development and deployment of AI systems is increasingly concentrated among a small number of well-resourced technology companies. This concentration stems not merely from access to capital and resulting computational infrastructure advantages, but also from asymmetric and unregulated access to training data.
While access to large-scale datasets is the main prerequisite of state-of-the-art language models, scholars and researchers have drawn attention to the importance of data quality in textual corpora used for AI training. Many have pointed to the need for curated, high-quality datasets, especially from library collections, which contain humanity’s most comprehensive and editorially refined record of knowledge, culture, and expression.
Currently, many academic researchers are denied access to this data for their own AI research due to a variety of legal, technical, and financial constraints. Our work on this project demonstrated a need for publicly accessible, research-oriented, computation-ready textual corpora to support academic work and non-profit AI development. The Public Interest Corpus initiative responds to this existing imbalance and pressing need by leveraging the unique position of research libraries to expand access to books data for academic and nonprofit AI training and computational research, thus ensuring that less-resourced institutions and individuals can gain equitable access to valuable data sources.

The report outlines our sense of how to move forward with books and AI, and seeks to address some hard issues that emerged from in-depth conversations we held, such as copyright questions and the needs of different users.

Given more recent technical developments, such as the Model Context Protocol, we also believe that the Public Interest Corpus will be able to serve not just noncommercial AI researchers, but also a broader audience among the public, students, and scholars. For instance, as I have noted in this space over the last year, AI shows great potential for creating a new digital entryway to the library, improving access and discovery by locating relevant books better than current library systems. For many, their interaction with AI will end after this phase of discovery and access; these library patrons will go on to read the books they have found rather than train new AI models with them. We should be enabling and encouraging these lighter uses of vectorized books as well as the heavier, more complex applications. Additional use cases emerged over the last year involving not one book or a million books, but collections at intermediate scales — what one might do with ten, a hundred, or a thousand books as part of a course, thesis, or research topic.

In the report, we also map out how the Public Interest Corpus should:

provide a secure technical environment for accessing data and provide the means to authenticate users and mitigate potentially infringing user behaviors
continually refine its data in order to increase the quality of the data we have about our books
encourage users to attribute books in their research through social and technical means
seek an environmentally sustainable infrastructure and mode of operations for its services

The team hopes that our report is a starting point rather than an endpoint, and we are currently working to make further progress toward implementation. My thanks again to the Mellon Foundation for generously supporting our work, and to Dave, Thomas, and Giulia, our helpful advisory board, and the many people we spoke to in 2025 for collaborating and advancing the Public Interest Corpus idea.

Open Knowledge Foundation: Learning together, building together – OKFN Newsletter April 2026

2026-04-29T18:03:02+00:00

AI Learning Labs, RightsCon, knowledge commons as critical digital infrastructure, Open Technology Research, and more.

The post Learning together, building together – OKFN Newsletter April 2026 first appeared on Open Knowledge Blog.

Open Knowledge Foundation: Join Open Knowledge at RightsCon 2026, Zambia

2026-04-28T18:41:32+00:00

Our team and network will discuss pressing issues such as reimagining AI, experiencing brain-computer interfaces, leveraging digital public goods, and more; check the full agenda

The post Join Open Knowledge at RightsCon 2026, Zambia first appeared on Open Knowledge Blog.

David Rosenthal: Dormant Digital Assets

2026-04-28T15:00:00+00:00

PsiQuantum's computer

Four and a half years ago I wrote The $65B Prize about the potential reward for developing a "sufficiently powerful quantum computer" capable of cracking Bitcoin's encryption. It was based on work by Aggarwal et al, who were then projecting it would happen between 2029 and 2044. The $65B was the notional value of the wallet containing the million Bitcoin Satoshi Nakamoto mined originally.i But I noted that:

Chainalysis estimates that about 20% of all Bitcoins have been "lost", or in other words are sitting in wallets whose keys are inaccessible. That is around another 3.6 million stranded Bitcoin or at the current "price" about $234B.

So the potential prize was almost $300B.

Nearly a year ago I followed up with The $740B Prize. There are two reasons why the prize was then bigger but is now smaller than that:

Bitcoin's "price" had then increased from about $65K to around $107K, but it is now around $76K.
Because the "market cap" of Michael Saylor's Strategy was 1.6 times the "market cap" of its stash of Bitcoin, it was possible to use Saylor's algorithm to amplify the prize. But the factor has decreased from 1.6 to 0.81, so the algorithm no longer works.

But the threat to Bitcoin, and other cryptocurrencies, is far worse than I described in either of these two posts. The date is closer and the range of threats much broader. Follow me below the fold for the details.

Ryan Babbush et 8 al's 57-page Securing Elliptic Curve Cryptocurrencies against Quantum Vulnerabilities: Resource Estimates and Mitigations is a comprehensive overview of, and an improvement to, the state of the art in Cryptographically Relevant Quantum Computers (CRQC), that is quantum computing applied to breaking the Elliptic Curve Discrete Logarithm Problem (ECDLP) underlying the cryptography used by most cryptocurrencies:

This whitepaper seeks to elucidate specific implications that the capabilities of developing quantum architectures have on blockchain vulnerabilities and potential mitigation strategies. First, we provide new resource estimates for breaking the 256-bit Elliptic Curve Discrete Logarithm Problem over the secp256k1 curve, the core of modern blockchain cryptography. We demonstrate that Shor’s algorithm for this problem can execute with either ≤ 1200 logical qubits and ≤ 90 million Toffoli gates or ≤ 1450 logical qubits and ≤ 70 million Toffoli gates. ... On superconducting architectures with 10⁻³ physical error rates and planar connectivity, those circuits can execute in minutes using fewer than half a million physical qubits. We introduce a critical distinction between “fast-clock” (such as superconducting and photonic) and “slow-clock” (such as neutral atom and ion trap) architectures. ... We survey major cryptocurrency vulnerabilities through this lens, identifying systemic risks associated with advanced features in some blockchains such as smart contracts, Proof-of-Stake consensus, and Data Availability Sampling mechanism, as well as the enduring concern of “abandoned” assets.

They identify three classes of attacks that such a CRQC would enable:

On-Spend Attacks: Attacks targeting transactions in transit. When a blockchain user broadcasts a transaction, an attacker must derive the private key within the window of time allowed before the transaction is recorded on the blockchain. This requires a quantum computer fast enough to solve ECDLP within the transaction settlement time of the target blockchain which ranges from hundreds of milliseconds to a few minutes (e.g., about 400 milliseconds for Solana, about 12 seconds for Ethereum, about 10 minutes on average for Bitcoin). On-spend attacks are also known as “short-range” or “just-in-time” attacks

At-Rest Attacks: Attacks targeting public keys that remain exposed onchain or offchain for long periods of time, such as dormant wallets with reused keys. The attacker has days (or more) to derive the private key. At-rest attacks are also known as “long-range” or “long-exposure” attacks

On-Setup Attacks: Attacks targeting fixed public protocol parameters that produce a universal reusable backdoor into a cryptographic protocol. The backdoor is created by means of a one-time off-line quantum computation on a CRQC and subsequent attacks utilizing it are executed on a classical computer. For example, an on-setup attack may involve the use of Shor’s algorithm to recover the so-called “toxic waste” discarded in a powers-of-tau trusted setup ceremony. While the Bitcoin blockchain is immune to on-setup attacks, some scaling solutions, such as Ethereum’s Data Availability Sampling mechanism, and privacy protocols, such as Tornado Cash, are vulnerable to this especially insidious attack mode.

Some quantum computer architctures will be capable of all three, but some will not be fast enough for On-Spend attacks:

The resource estimates we describe below indicate that superconducting, photonic, and silicon spin qubit CRQCs, with their fast gates and short quantum error correction cycles, will be able to solve ECDLP in the span of a few minutes and thus, to launch on-spend attacks. By contrast, the elementary operations on neutral atom and ion trap devices are about two to three orders of magnitude slower. As a consequence, we do not expect CRQCs in these slower architectures to be able to launch on-spend attacks. We will refer to the former as fast-clock CRQCs and to the latter as slow-clock CRQCs

Babbush et al Fig. 1

There have been major improvements in both hardware and software since the previous estimates. In particular, software:

We are reporting here that our team has developed logical circuits to break ECDLP on elliptic curves over finite fields with n-bit prime modulus and n-bit group order requiring approximately 4.5n space. ... At n = 256 bits, the circuits use either 1200 logical qubits and 90 million Toffoli gates or 1450 logical qubits and 70 million Toffoli gates. In terms of the spacetime volume (a key resource which in particular drives the quantum error correction overhead), these estimates represent roughly an order of magnitude improvement over the most efficient prior work when applied to a single ECDLP instance. ... Our findings apply directly to ECDLP on secp256k1 — an elliptic curve widely used in digital signatures on popular blockchains, such as Bitcoin and Ethereum.

And hardware design:

The physical resource estimates we have discussed here (e.g., half a million physical qubits) assume relatively benign hardware capabilities, such as a planar architecture with degree-four connectivity and 10−3 physical gate error rates (i.e., consistent with a scaled up version of Google’s quantum processors that have been demonstrated experimentally). More aggressive hardware assumptions — such as the “bicycle” architecture used for 2-gross qLDPC codes — could drop qubit counts closer to one hundred thousand physical qubits, but this approach requires non-local degree-seven connectivity that has yet to be demonstrated in actual superconducting qubit devices.

There are now many companies trying to turn the designs into working fast- and slow-clock hardware:

Google Quantum AI, IBM Quantum, Amazon, D-Wave, Rigetti and IQM are developing superconducting qubit architectures; PsiQuantum and Xanadu are building photonic quantum computers while Diraq and Intel are working on spin qubit devices. ... Simultaneously, many companies, including IonQ, Quantinuum (a subsidiary of Honeywell) and Alpine Quantum Technologies are pursuing ion trap quantum processors while others, such as QuEra, Infleqtion, Atom Computing, Pasqal, and Logiqal are developing neutral atom devices.

Babbush et al thus argue that, if the first CRQC is fast-clock, all three attack types will arrive simultaneously:

These facts imply that a superconducting CRQC capable of performing at-rest attacks against static holdings recorded on the blockchain would likely also be capable of executing on-spend attacks against active transactions. As we discuss in more detail later on, we do not expect meaningful scaling challenges between a quantum computer with 1200 logical qubits and one with 1450, so, in order to focus and simplify subsequent discussion, we assume that first-generation fast-clock CRQCs may be able to solve ECDLP on secp256k1 and similar elliptic curves in about 9 minutes on average.

A major problem with current techniques for stealing cryptocurrency is that the proceeds need to be rapidly laundered because the thefts are detectable. But if the contents of a vulnerable wallet move to an invulnerable one, it is likely that the "owner" of the private key was taking a sensible precaution, not that some CRQC cracked the key. This is especially true of dormant assets; no-one is watching the wallet.

Although the paper's analysis of On-Spend and Setup attacks is fascinating and important, much of this post will focus on the At-Rest attacks on Bitcoin that my previous posts discussed. Babbush et al summarize the problem:

Dormant digital assets, including those abandoned or inaccessible due to lost private keys, pose a distinct and critical challenge. We highlight the example of Bitcoin’s Pay-to-Public-Key (P2PK) locking scripts, which secure over 1.7 million BTC. The total amount of dormant quantum-vulnerable bitcoin may reach 2.3 million BTC when all script types are considered. Unlike active wallets that can migrate to new standards, dormant assets cannot be “fixed” via forks that enable PQC protocols for future transactions. They represent a fixed target — tens or hundreds of billions of dollars in value that will eventually become accessible to a quantum attacker. The community will soon face difficult, unprecedented decisions regarding the fate of these assets, forcing tradeoffs between the immutability of cryptographic property rights and the economic stability of the network.

Babbush et al Fig. 4

Bitcoin wallets are vulnerable to an At-Rest attack if their public ECDSA key is visible on the blockchain. Over time, the way that transactions are encoded on the blockchain, the "scripts", have evolved. Babbush et al's Figure 4 shows this evolution.

A transaction contains an unlocking script, proving that the private key owns the wallet, and a locking script that transfers coins to the recipient. Some script types reveal the public key and are thus vulnerable to an At-Rest attack, some reveal only its hash and are thus immune unless the script is re-used.

Babbush et al Fig. 5

Babbush et al's Figure 5 shows the numbers of Bitcoin secured by the various script types. The shaded areas represent the script types that are vulnerable to At-Rest attacks as soon as any type of CRQC exists. A little over 1.7M BTC (~$130B) are vulnerable even if the script has not been re-used, around another 5.2M BTC are vulnerable if the script has been re-used. Thus the total at risk is currently "worth" around $525B.

As I have been writing, the part of the problem that cannot be solved by upgrading to post-quantum cryptography is what Babbush et al call "Dormant Digital Assets":

Inevitably, some vulnerable assets will not migrate to post-quantum protocols in time or possibly ever, perhaps because their owners do not learn of the threat until it is too late or perhaps because they have lost their private keys. The Ethereum blockchain’s contract accounts present similar long-tail migration issues. Thus, in addition to planning and executing upgrades to cryptographic protocols, each cryptocurrency community also faces challenges regarding quantum-vulnerable assets and smart contracts that may linger on public blockchains for an extended or indefinite period of time.

Despite lack of unambiguous precedent, many jurisdictions could classify accessing abandoned cryptographic assets, such as the P2PK coins, without authorization as theft. However, we maintain that if protocol changes are not made, vulnerable assets will eventually be cracked by quantum computers and taken irrespective of the law. In the absence of a clear resolution, these assets are likely to become a lucrative target for bad actors. We quantify the scale of some of the dormant assets at stake in Figure 13.

Babbush et al Fig. 13

After all, "code is law". The total is about 2.3M BTC "worth" about $175B. It might take months (fast-clock) or years (slow-clock) for a single CRQC to compromise the wallets with the 1.7M BTC. Of course, the attackers would choose the wealthiest wallets first, working left-to-right across Figure 13, and there is no reason to assume that they would only have a single CRQC, so the bulk of the loss would happen more quickly.

Bitcoiners have identified three responses to the problem that they could take, if it were possible to achieve consensus on which:

Do Nothing: accept that the 2.,3M BTC would be stolen and become part of the circulating supply, thus putting downward pressure on the "price".
Burn: implement a soft-fork that renders the content of vulnerable wallets unspendable after a certain date. Provided the date is before the first CRQC, this removes them from the circulating supply and avoids downward pressure on the "price". It does conform to the "not your keys, not your coins" mantra.
Hourglass: accept that the 2.3M BTC would be stolen but mitigate the effect on the "price" by limiting the rate at which these assets could be spent and thus enter the circulating supply.

As usual, concensus in the Bitcoin community is likely to be hard to achieve, giving an advantage to Do Nothing:

Those who consider digital property rights fundamental tend to have strong objections to the Burn proposal. Large Bitcoin holders are likely concerned about a potential supply shock and its effect on Bitcoin price. Miners may welcome Do Nothing and Hourglass proposals due to potential increase in transaction fees and volumes. The diversity and complexity of the Bitcoin community makes the ultimate outcome of these ongoing debates hard to predict. Indeed, an informal poll at 2025 Presidio Bitcoin Quantum Summit in San Francisco saw roughly equal support for each of the three categories of solutions.

Babbush et al add a fourth option, one or more sidechains to which public-spirited CRQC operators could, for a fee, send the contents of wallets they compromised, and other sidechains holding holding cryptographic proofs of ownership of the dormant assets in question. These sidechains would form a somewhat complex and thus risky ecosystem, and would be costly. Operating an early CRQC will be expensiuve, so the public-spritied operators would need to charge significant fees. Figure 13 shows that the bulk of the value is in the first 1000 wallets, so it is likely that this solution would leave something like 100K dormant wallets uneconomic to compromise.

The authors also review threats to other cryptocurrencies. One obvious threat is:

the objective of preserving transaction confidentiality on privacy-preserving blockchains, such as Zcash and Monero, cannot be fully achieved due to retroactive degradation of ECDLP-protected privacy of known addresses by quantum-capable adversaries.

Ethereum is by far the largest of the systems that are already taking proactive quantum-proofing steps. This is important because Etereum is far more exposed than Bitcoin to quantum attacks, as Babbush et al recount:

The account model uses vulnerable elliptic curves as a core component of onchain identity, putting all accounts that have carried out transactions at risk including high value accounts, such as exchange hot wallets. Smart contracts with exposed admin keys that cannot be easily rotated (without draining and replacing the contracts themselves) create a logistical bottleneck for security upgrades that puts “low ether, high leverage” accounts and contracts responsible for tokenized real-world assets, oracles, bridges, guardians, etc. at risk. Moreover, the potential compromise of validators threatens the integrity of the Proof-of-Stake consensus mechanism itself, creating an existential risk to the chain’s continued operation. Finally, the vulnerability of Data Availability Sampling mechanism opens it up to on-setup attacks that can be launched without a quantum computer using a reusable exploit created once on a CRQC.

For example, Tornado Cash is a smart contract whose administrative public key is 0x0000, which indicates that adminstrative control has been relinquished. The Presumably it will continue to function unless and until Ethereum decides to stop executing contracts with this key. Or, possiby, a CRQC could find the priovate key for 0x0000. Tornado's wallets have exposed keys, so could be drained unless each user removes their funds before the attack.

Babbush et al have a section addressing Public Policy Options for the Challenge of Dormant Assets. They start by arguing that government action to protect the "price" of BTC and similar cryptocurrencies by mandating the Burn option would be highly unlikely to succeed. They then argue that one approach would be to use existing laws on lost, abandoned and unclaimed assets:

if an owner of dormant coins has known for years that their assets are at risk and has failed to transact them to a post-quantum address, then they may be deemed to have failed to assert their rights through inaction.

But they point out many difficulties with this approach. For example, in the US the Revised Uniform Unclaimed Property Act (2016) is a model for relevant state laws on abandoned property, allowing it to be transferred to the custody of the state. The law assumes that the assets are in the custody of a "holder", a business such as a bank, but:

no party involved in the operation of the Bitcoin blockchain clearly meets the legal requirements to be the “holder” of the dormant coins. Indeed, none of them possess or control the coins since none of them know the private key.

They also discuss the:

spectre of dormant assets falling to rogue actors as a national security risk

and suggest that some governments will decide to use CRQCs to grab and maybe burn dormant assets. What they mean is the US is worried that the North Koreans might acquire a CRQC. They are the masters of stealing cryptocurrency via conventional techniques, as we apparently see with the recent compromise of Kelp DAO.

As regards Bitcoin, the authors recommend that governments establish a legal framework for dormant digital assets similar to that for conventional abandoned assets, and that the Bitcoin community decide to implement the Burn option. Given the current difficulty of passing stablecoin legislation and the history of consensus in the Bitcoin community, I would expect that neither will happen in time.

As regards Ethereum, the more sophisticated technology and governance, combined with the absolutely catastrophic effects a CRQC could have on the ecosystem, give some confidence that timely mitigations are possible.

HangingTogether: New OCLC Research Report: The Library Beyond the Library

2026-04-27T14:48:35+00:00

Having recently released a Data Insights briefing on the Italian presence in the global published record, I’m inspired to introduce our latest OCLC Research report with a quote from Machiavelli:

Men in general judge more from appearances than from reality. All men have eyes, but few have the gift of penetration. (Niccolò Machiavelli, The Prince)

What’s the connection? Read on …

Our new report, The Library Beyond the Library: Recasting the Library Value Proposition for Visibility and Impact, begins with the observation that academic libraries are taking on important new roles throughout the research lifecycle: from publishing, to research data management, to impact assessment. In doing so, their value proposition to the rest of the institution is evolving, and at the same time, becoming more complex and potentially more opaque to campus stakeholders.

Libraries have a long-standing, well-understood value proposition centered around collections—a perception that has persisted even as libraries have developed new offerings across a wide range of emerging areas of research support. Fixed ideas about library roles and impact create a challenge: despite significant library investments, institutional stakeholders often don’t understand, recognize, or are simply unaware of library service offerings in these new areas. Instead, the traditional collections-centric view of the library endures.

Collection stewardship remains a vital aspect of the library mission, but increasingly, academic libraries face a disconnect between their evolving services and institutional perceptions. As Machiavelli observes, appearances often overshadow reality. Academic libraries offer valuable capacities and expertise well-calibrated to meet institutional research needs and priorities, but the perception that library impact is limited to its traditional role of collections steward will nonetheless prevail if nothing is done to correct it.

This has real consequences. Perceptions of the library’s value proposition based on fixed ideas of its role and impact make it difficult for the true scope of library capacities and expertise—the “reality”, as Machiavelli expressed it—to filter through to institutional stakeholders. This makes it hard for the library to get a seat at the table for institution-wide discussions and policy-making on topics like data governance, research metrics, or open research practices; moreover, it can lead to diminished influence, and ultimately, reduced funding.

This is a challenge, but also an opportunity for libraries to clarify their continued relevance to the institutional research enterprise. And that brings us back to OCLC Research’s new report, The Library Beyond the Library: Recasting the Library Value Proposition for Visibility and Impact. Based on in-depth interviews with international research library leaders, desk research, and accumulated insights from our previous studies of research support services, The Library Beyond the Library helps libraries navigate these trends by providing a framework and insights that support strategic planning aimed at elevating the library’s visibility and impact within its parent institution.

While the report findings were derived in the context of research support services in academic libraries, we believe they apply equally to many other areas of strategic importance to academic libraries, such as institutional priorities for student success, as well as to public and other types of libraries.

What do we mean by “the library beyond the library”? It’s an operational principle that emphasizes engagement “beyond the library” with the broader institutional environment, in support of the institutional research and learning mission. In our report, we argue that this operational principle increasingly shapes libraries’ ability to fulfill their mission, retain influence, and demonstrate impact and value.

The library beyond the library principle focuses on engagement through three channels:

Strategic Alignment—Aligning library priorities with institutional goals
Collaboration—Partnering with other institutional units to advance shared priorities
Storytelling—Communicating the library’s evolving value proposition to stakeholders

These channels of engagement translate into important strategic questions for libraries:

How can library services and expertise support, or in some cases evolve in response to, institutional priorities?
What partnership opportunities exist with other institutional units, and how can libraries structure them effectively?
How can libraries construct and communicate compelling narratives about their evolving value and impact to stakeholders?

The reality of the academic library on campus has expanded well beyond its traditional role of collections steward, but its appearance to many stakeholders—its perceived value proposition—often has not kept pace. This creates risks, because visibility combined with a clear stakeholder understanding of impact drives influence, inclusion in institutional decision-making, and funding.

“The library beyond the library“ is not a slogan, but a practical response to mitigate these risks as part of a process of updating and communicating the library’s evolving value proposition. By investing in intentional strategic action across all three framework channels—strategic alignment (tying services and expertise to institutional priorities), collaboration (building relationships and shared commitments with key units), and storytelling (making impact clear to stakeholders)—the library can demonstrate it is and will continue to be a dynamic partner in shaping the future of scholarship and research at its parent institution.

We invite you to read the Library Beyond the Library report and consider how you might use the framework at your institution to assess current services and expertise, identify cross-institutional partnership opportunities that showcase library capacities, and reimagine narratives about the library value proposition.

The post New OCLC Research Report: The Library Beyond the Library appeared first on Hanging Together.

Ed Summers: Weekly Bookmarks

2026-04-26T04:00:00+00:00

These are some things I’ve wandered across on the web this week.

🔖 Talk Talk - Live at Montreux 1986

From Aquarium Drunkard:

🔖 The West Forgot How to Make Things. Now It’s Forgetting How to Code

🔖 There Will Be a Scientific Theory of Deep Learning

🔖 EHRAG: Bridging Semantic Gaps in Lightweight GraphRAG via Hybrid Hypergraph Construction and Retrieval

Graph-based Retrieval-Augmented Generation (GraphRAG) enhances LLMs by structuring corpus into graphs to facilitate multi-hop reasoning. While recent lightweight approaches reduce indexing costs by leveraging Named Entity Recognition (NER), they rely strictly on structural co-occurrence, failing to capture latent semantic connections between disjoint entities. To address this, we propose EHRAG, a lightweight RAG framework that constructs a hypergraph capturing both structure and semantic level relationships, employing a hybrid structural-semantic retrieval mechanism. Specifically, EHRAG constructs structural hyperedges based on sentence-level co-occurrence with lightweight entity extraction and semantic hyperedges by clustering entity text embeddings, ensuring the hypergraph encompasses both structural and semantic information. For retrieval, EHRAG performs a structure-semantic hybrid diffusion with topic-aware scoring and personalized pagerank (PPR) refinement to identify the top-k relevant documents. Experiments on four datasets show that EHRAG outperforms state-of-the-art baselines while maintaining linear indexing complexity and zero token consumption for construction.

🔖 The Night Manager (British TV series)

The Night Manager is a British spy thriller television serial based on the 1993 novel by John le Carré and adapted by David Farr. The six-part first series, directed by Susanne Bier and starring Tom Hiddleston, Hugh Laurie, Olivia Colman, Tom Hollander, David Harewood and Elizabeth Debicki, began broadcasting on BBC One on 21 February 2016.

🔖 What Will It Take to Get A.I. Out of Schools?

Immordino-Yang told me that the ultimate goal of any school assignment is not the finished project itself but the experience of having done it—an experience that A.I. tools are intended to abbreviate or obviate. With their prettifying intrusions and impatient, lurking presence, they block and reroute a young person’s natural, gradual progression toward cognitive maturity, “especially one who is still developing the neuropsychological substrate for creating narratives and thinking through arguments over time,” Immordino-Yang said. “It’s a fragile process, and it’s being interrupted.”

🔖 iocaine

This software is not made for making the Crawlers go away. It is an aggressive defense mechanism that tries its best to take the blunt of the assault, serve them garbage, and keep them off of upstream resources. Even though a lot of work went into making iocaine efficient, and nigh invisible for the legit visitor, it is an aggressive defender nevertheless, and will require a few resources - a whole lot less than if you’d let the Crawlers run rampant, though.

Before you deploy it, be sure you understand that iocaine does not make the bots go away. It tries to poison them, so they’d go away forever in the long run. If you’re looking for a way to return the favour, to “reward” these crawlers for their relentless assault, this is the tool you’re looking for.

🔖 AI as a Fascist Artifact

“AI” is being introduced increasingly into government processes: “AI” is promised to bring more efficiency into the administration, is supposed to “reduce bureaucracy”. But bureaucracy is not just an annoyance but one of the central tools that democratic societies have established to realize the core idea of democracy: Transparency in the application of power in order to be able to control said power. Democracy is not just about voting but about ensuring that all power – especially by the state – is used in accordance with the law and in a fair way. Stochastic “AI” systems break that promise. The “AI” just says that you do not get the support you need. No idea why, might be a bug or a deeply racist training data set or something else. Nobody knows. Now it is on you to prove that you are in the right, it is on you to fight for your right because the processes that were supposed to protect your rights are hollowed out in order to make them faster: We are forcing marginalized, disenfranchised people to fight against a black box trained on the data that already contains their disenfranchisement

🔖 You’re about to feel the AI money squeeze

Investors have poured hundreds of billions of dollars into companies like OpenAI and Anthropic to help them scale and build out their compute. Now, they’re expecting returns. After years of offering cheap or totally free access to advanced AI systems, the bill is starting to come due — and downstream, users are beginning to feel the pinch.

🔖 RustFS

Instantly replace MinIO & S3. Zero GC, maximum throughput.

🔖 Finishing Things

Exactly! Left behind? You can’t leave me behind fast enough. I’ve never wanted to be left behind so bad in my life. I’m utterly incapable of FOMO about this stuff. Do I have vague future concerns about my career and what must (surely) be a coming economic crash? Sure, but there is absolutely nothing that has convinced me that I’m “missing out” on anything.

At this point, all efforts by boosters and sloppologists just make me feel more defiant.

🔖 Go Ahead And Use AI. It Will Only Help Me Dominate You.

The tepid, conformist nature of your AI-assisted prose will only make my unexpected bons mots stand out more sharply. While you lean on a technological crutch of grammatical mediocrity to drag your essays over the finish line, I’ll be metaphorically zipping past you on my “magic carpet” of words emerging directly from my own declining and unpredictable brain. Over time, the intellectual box into which AI has seduced your creative process will suffocate you, leaving your bereft readers little choice but to drift into my subscription base.

🔖 Friends Don’t Let Friends Use Ollama

Ollama gained traction by being the first easy llama.cpp wrapper, then spent years dodging attribution, misleading users, and pivoting to cloud, all while riding VC money earned on someone else’s engine. Here’s the full history, and why the alternatives are better

David Rosenthal: Metastablecoin Fragmentation (updated)

2026-04-24T18:13:50+00:00

A fundamental problem for decentralized systems like permissionless blockchains is that their security depends upon the cost of an attack being greater than the potential reward from it. Various techniques are used to impose these costs, generally either Proof-of-Work (PoW) or Proof-of-Stake (PoS). These costs have implications for the economics (or tokenomics) of such systems, for example that their security is linear in cost, whereas centralized systems can use techniques such as encryption to achieve security exponential in cost.

Shin Figure 3

Now, via Toby Nangle's Stablecoin = Fracturedcoin we find Tokenomics and blockchain fragmentation by Hyun Song Shin, whose basic point is that these costs must be borne by the users of the system. For cryptocurrencies, this means through either or both transaction fees or inflation of the currency. The tradeoff between cost and security means that there is a market for competing blockchains making different tradeoffs. In practice we see a vast number of competing blockchains:

Tether’s USDT sits on 107 different ledgers. ... USDC sits on 125.

The chart shows Ethereum losing market share against competing blockchains.

Shin's analysis uses game theory to explain why this fragmentation is an inevitable result of tokenomics. Below the fold I go into the background and the details of Shin's explanation.

Background

In 2018's Cryptocurrencies Have Limits I discussed Eric Budish's The Economic Limits Of Bitcoin And The Blockchain, an important analysis of the economics of two kinds of "51% attack" on Bitcoin and other cryptocurrencies based on PoW blockchains. Among other things, Budish shows that, for safety, the value of transactions in a block must be low relative to the fees in the block plus the reward for mining the block.

In 2019's The Economics Of Bitcoin Transactions I discussed Raphael Auer's Beyond the doomsday economics of “proof-of-work” in cryptocurrencies, in which Auer shows that:

proof-of-work can only achieve payment security if mining income is high, but the transaction market cannot generate an adequate level of income. ... the economic design of the transaction market fails to generate high enough fees.

Source

Bitcoin's costs are defrayed almost entirely by inflating the currency, as shown in this chart of the last year's income for miners. Notice that the fees are barely visible.

It has been known for at least a decade that Bitcoin's plan to phase out the inflation of the currency was problematic. In 2024's Fee-Only Bitcoin I wrote:

In 2016 Arvind Narayanan's group at Princeton published a related instability in Carlsten et al's On the instability of bitcoin without the block reward. Narayanan summarized the paper in a blog post:

Our key insight is that with only transaction fees, the variance of the miner reward is very high due to the randomness of the block arrival time, and it becomes attractive to fork a “wealthy” block to “steal” the rewards therein.

So Bitcoin's security depends upon the "price" rising enough to counteract the four-yearly halvings of the block reward. In that post I made a thought-experiment:

As I write the average fee per transaction is $3.21 while the average cost (reward plus fee) is $65.72, so transactions are 95% subsidized by inflating the currency. Over time, miners reap about 1.5% of the transaction volume. The miners' daily income is around $30M, below average. This is about 2.5E-5 of BTC's "market cap".

Lets assume, optimistically, that this below average daily fraction of the "market cap" is sufficient to deter attacks and examine what might happen in 2036 after 3 more halvings. The block reward will be 0.39BTC. Lets work in 2024 dollars and assume that the BTC "price" exceeds inflation by 3.5%, so in 12 years BTC will be around $98.2K.

To maintain deterrence miners' daily income will need to be about $50M, Each day there will be about 144 blocks generating 56.16BTC or about $5.5M, which is 11% of the required miners' income. Instead of 5% of the income, fees will need to cover 89% of it. The daily fees will need to be $44.5M. Bitcoin's blockchain averages around 500K transactions/day, so the average transaction fee will need to be around $90, or around 30 times the current fee.

Average fee/transaction

Bitcoin users set the fee they pay for their transaction. In effect they are bidding in a blind auction for the limited supply of transaction slots. Miners are motivated to include high-fee transactions in their next block. If there were an infinite supply of transactions slots miners' fee income would be zero. In practice, much of the timethe supply of slots exceeds demand and fees are low. At times when everyone wants to transact, such as when the "price" crashes, the average fee spikes enormously.

There was thus a need for a consensus mechanism that did not depend upon inflation. In 2020's Economic Limits Of Proof-of-Stake Blockchains I discussed a post entitled More (or less) economic limits of the blockchain by Joshua Gans and Neil Gandal in which they summarize their paper with the same title. The importance of this paper is that it extends the economic analysis of Budish to PoS blockchains. Their abstract reads:

Cryptocurrencies such as Bitcoin rely on a ‘proof of work’ scheme to allow nodes in the network to ‘agree’ to append a block of transactions to the blockchain, but this scheme requires real resources (a cost) from the node. This column examines an alternative consensus mechanism in the form of proof-of-stake protocols. It finds that an economically sustainable network will involve the same cost, regardless of whether it is proof of work or proof of stake. It also suggests that permissioned networks will not be able to economise on costs relative to permissionless networks.

Source

In 2022 Ethereum switched from Proof-of-Work to Proof-of-Stake, reducing its energy consumption by around 99%. This chart shows that, like Bitcoin, until the "Merge" the costs were largely defrayed by inflating the currency. After the "Merge" the blockchain has been running on transaction fees.

Shin's Analysis

Here is a summary of Shin's analysis.

Notation

There is a continuum of validators i.
For validator i ∈ [0;1], the cost of contributing to governance is c_i > 0.
The blockchain needs at least a fraction k̂ of the validators contributing to be secure. Shin writes:

There are two special cases of note: k̂ = 1 (unanimity, corresponding to full decentralisation where every validator must participate for the blockchain to function) and k̂ = 0 which corresponds to full centralisation, where one validator has authority to update the ledger.
k̂ = 1 is impractical,lacking fault tolerance. k̂ = 0 is much more practical, it is the traditional trusted intermediary.
If the blockchain is secure, each contributing validator earns a reward p > 0. A non-contributing validator earns zero.
The validators share a common cost threshold c^*. If c_i < c^*, validator i contributes, if c_i > c^* validator i does not.

Argument

Each validator will want to contribute only if at least k̂ - 1 other validators contribute, which poses a coordination problem. The case of particular interest is the validator with c_i = c^*. Shin writes:

Intuitively, even though the marginal validator may have very precise information about the common cost c^*, the validator faces irreducible uncertainty about how many other validators will choose to contribute. It is this strategic uncertainty — uncertainty about others' actions — that is the central feature of the coordination problem.

This "strategic uncertainty" is similar to the attacker's uncertainty about other peers' actions that is at the heart of the defenses of the LOCKSS system in our 2003 paper Preserving peer replicas by rate-limited sampled voting.

Shin Figure 6

Because the marginal validator's c_i = c^*, the decision whether or not to contribute makes no difference. Sin's Figure 6 explains this graphically. Rectangle A is the loss if k < k̂ and rectangle B is the gain if k > k̂. Setting them equal gives:

c^*k̂ = (p - c^*)(1 - k̂)

which simplifies to:

c^* = p(1 - k̂)

Shin and Morris earlier showed that this is the unique equilibrium no matter what strategy the validators use.

Result

What this means is that successful validation depends upon the reward p being large enough so that:

p ≥ c 1 − κ̂

Shin writes:

Note that the required reward p explodes as k̂ → 1. This is the central result of the paper: the more decentralised the blockchain (the higher the supermajority threshold), the higher must be the rents that accrue to validators. In the limiting case of unanimity (k̂ = 1), no finite reward can sustain the coordination equilibrium.

Shin Figure 1

This yet another result showing that a reasonably secure blockchain is unreasonably expensive. The complication is that, much of the time, transactions are cheap because the demand for them is low. Thus most of the time validators are not earning enough for the risks they run. But:

When many users want to transact at the same time, they bid against each other for limited block space, and fees spike — much as taxi fares surge during rush hour. Figure 1 shows how Ethereum gas fees exhibited sharp spikes during periods of network congestion, such as during surges in decentralised finance (DeFi) activity or spikes in the minting of non-fungible tokens (NFTs). These spikes are not merely a reáection of excess demand; they are the mechanism through which the blockchain extracts the rents needed to sustain validator coordination.

Note that these spikes mean that the majority of the time fees are low but the majority of transactions face high fees. It is this "user experience" that drives the fragmentation that Shin describes:

When demand for block space is high, fees rise and validators are well compensated. But high fees deter users, especially those making small or routine transactions. These users are the first to migrate to competing blockchains that offer lower fees — blockchains that can offer lower fees precisely because they have lower coordination thresholds (and hence less security). The users who remain on the more secure blockchain are those with the highest willingness to pay: institutions, large DeFi protocols, and transactions where security and censorship resistance are paramount. This sorting of users across blockchains is the essence of fragmentation.

Shin notes that:

The fragmentation argument is the flipside of blockchain's "scalability trilemma," as described by Vitalik Buterin, who posed the problem as the impossibility of attaining, simultaneously, a ledger that is decentralised, secure, and scalable.

Source

It is worth noting that Buterin's trilemma is a version for PoS of the trilemma Markus K Brunnermeier and Joseph Abadi introduced for PoW in 2018's The economics of blockchains. See The Blockchain Trilemma for details.

Shin's focus is primarily on the effects of fragmentation on stablecoins. He notes that:

Rather than converging on a single platform, stablecoin activity is scattered across many chains (Figure 4). As of late 2025, Ethereum held the majority of total stablecoin supply but was facing competition from Tron and Solana, each of which had attracted tens of billions of dollars in stablecoin balances. Each chain serves different geographies and use cases: Ethereum for institutional settlement, Tron for low-cost remittances, Solana for retail payments and DeFi activity.

This fragmentation among blockchains would not matter much if stablecoins were interoperable between them, but they are confined to the blockchain on which they were minted:

A USDC token on Ethereum is not the same as a USDC token on Solana — they exist on separate ledgers that have no native way of communicating with each other. Transferring between chains requires the use of bridges: specialised software protocols that lock tokens on one chain and issue equivalent tokens on another. These bridges introduce additional risks, including vulnerabilities in the smart contract code — bridge exploits have accounted for billions of dollars in cumulative losses — and they impose costs and delays that undermine the seamless transferability that is the hallmark of money. The result is a landscape in which stablecoins from the same issuer exist in multiple, non-fungible forms across different blockchains, fragmenting liquidity and undercutting the network effects that should be the strength of a widely adopted payment instrument.

Discussion

As I've been pointing out since 2014, very powerful economic forces mean that Decentralized Systems Aren't. So the users paying for the more expensive transactions because they believe in decentralization aren't getting what they pay for.

Source

As I wrote in 2024's It Was Ten Years Ago Today:

The insight applies to Proof Of Stake networks at two levels:

Block production: over the last month almost half of all blocks have been produced by beaverbuild.
Staking: Yueqi Yang noted that:

Coinbase Global Inc. is already the second-largest validator ... controlling about 14% of staked Ether. The top provider, Lido, controls 31.7% of the staked tokens,
That is 45.7% of the total staked controlled by the top two.

Source
In addition all these networks lack software diversity. For example, as I write the top two Ethereum consensus clients have nearly 70% market share, and the top two execution clients have 82% market share.

Shin writes as if more decentralization equals more security even though it doesn't happen in practice, but this isn't really a problem. What the users paying the higher fees want is more security, and they are probably getting because they are paying higher fees. As I discussed in Sabotaging Bitcoin, the reason major blockchains like Bitcoin and Ethereum don't get attacked is not because the (short-term) rewards for an attack are less than the cost. It is rather that everyone capable of mounting an attack is making so much money that:

those who could kill the golden goose don't want to.

Shin Figure 3

In any case what matters for Shin's analysis isn't that the users actually get more security for higher fees, but that they believe they do. Like so much in the cryptocurrency world, what matters is gaslighting. But what the chart showing Ethereum losing market share shows is that security is not a concern for a typical user.

Update 22^nd April 2026

Shin warned about the risks introduced by bridges:

"These bridges introduce additional risks, including vulnerabilities in the smart contract code — bridge exploits have accounted for billions of dollars in cumulative losses

Sidhartha Shukla covered the latest example of what Shin was warning about in a trio af articles. First, Crypto Hack Worth $290 Million Triggers DeFi Contagion Shock:

Hackers exploited a cross-chain bridge on Saturday, draining nearly $300 million from a key piece of decentralized finance infrastructure and setting off a ripple effect across multiple crypto platforms.

The attacker siphoned about 116,500 rsETH — a token issued by Kelp DAO that represents “restaked” Ether — by targeting a bridge built using LayerZero, a system that allows different blockchains to communicate. The total losses are estimated at roughly $293 million, making it the largest DeFi exploit of 2026.

Siphoned isn't quite the right word. The attackers were able to mint tokens purpoting to represent staked ETH, but the ETH didn't exist. Second, Crypto Hack Sparks $9 Billion Outflows From Top DeFi Lender":

The hackers deposited about $200 million of the tokens they stole on Aave as collateral for borrowing another cryptocurrency, according to cybersecurity researcher PeckShield. That move sparked fears among depositors about possibly worthless collateral on Aave, causing a rush for the exit, crypto portfolio manager Pratik Kala said.

All told, Aave has recorded some $9 billion of net outflows since Saturday, when news of the heist first emerged, data from industry tracker DefiLlama shows. Total value locked on the platform — a measure of assets held there — plunged by more than a third to $17.5 billion.

It is the equivalent of a bank run caused by missing or devalued collateral. Third, Hackers Behind $300 Million Crypto Theft Now Laundering Loot:

Wallets linked to the roughly $300 million hack of Kelp DAO — a decentralized-finance protocol — have begun moving funds through services designed to obscure their trail, according to blockchain security firm Cyvers. About $175 million worth of stolen assets was shifted into two new wallets and is being routed through platforms including THORChain, Umbra and BitTorrent, Cyvers said.

The activity picked up on Tuesday, shortly after Arbitrum, a network running on the Ethereum blockchain, froze around $75 million of the stolen assets. Arbitrum described the measure as an emergency action taken following input from law enforcement.

This illustrates a fundamental problem with "tokenization", which was once the next big thing. The existence of the backing depends upon the security of the mint. This must be a "smart contract" with (a) no bugs and (b) secure private key(s) owned by (c) humans immune from social engineering. What could possibly go wrong?

Update 24^th April 2026

I just figured out an implication of Shin's fragmentation mechanism. Suppose you are a stablecoin user on blockchain A. You encounter a fee spike and decide to move to cheaper, less secure blockchain B. You need to pay a bridge to move your stablecoins to blockchain B, but you decide this investment in cheaper transactions is worth it. You are never going back to blockchain A, not just because it is too expensive, but also because you now have sunk costs of the move, and also because you would have to pay the bridge to move back. So Shin's mechanism implements a ratchet; users will flow from more expensive to cheaper blockchains but will not flow in the opposite direction. Thus the security afforded the median user will decay over time.

Library | Ruth Kitchin Tillman: Open Refine: Repeating Groups of Operations (the Easy Way)

2026-04-23T00:00:00+00:00

I’m currently working on a project which involves a lot of spreadsheets exported from our ILS using the same process. To get started with each in OpenRefine, I load the spreadsheet as a project and apply the same 7 manipulations, including the steps for blanking down within records I previously described. These are OpenRefine-specific, not something I could do during export or before adding to OpenRefine.

After doing the same 7 operations on the 20th spreadsheet, I started wondering about reproducibility. Surely, there’s a way to do the same set of actions without all those clicks?

This question led me on a 3-hour journey learning how to manipulate projects with the API… only to realize in the process that there was a much, much simpler way. Today I’ll be sharing the simpler way, but I will write a post about the API, since when I brought it up in a group chat with plenty of regular OpenRefine users, no one else was familiar with it either.

Short and Sweet: Reusing Operations in OpenRefine

OpenRefine logs all of your actions in the “Undo/Redo” tab on the left sidebar. Despite visiting this area many times to undo and redo my actions, I had never experimented the “Extract” or “Apply” buttons above the list of operations performed. They’d faded into the background.

Yeah, it turns out those are really useful.

When you click Extract, you’re presented with some options: you can select which functions you want to export, see and copy the code, or export it to a text file. Apply is the inverse: just paste in the code or upload an exported text file and run the result.

For a lot of power users, this info may be enough to get you oriented and experimenting. But if you’re interested in more (we’re still on the simple stuff) or want a walkthrough with screenshots and sample code, carry on!

Operational Background

For some reference, I’m going to describe the data I have and the operations I was performing. I’ll also include some sample data for you to practice with below.

My spreadsheets have the following columns:

Catalog Key
Title
856
Barcode
Item Type
Library
Home Location

They are based around the Item represented in the second half, so the first half just repeats bibliographic data for each entry. As described in the previous post I needed to get this data into the shape of an OpenRefine record by “blanking down” the Catalog Key. Then, for easier visual processing, I did the same to the Title and 856 fields.

Multi-value 856s are represented by multi-line entries in the cell, so I can blank down the field without losing any data. So I wanted to apply cell transformations to both those fields to make them unique within a record. Then blank that down. Then re-transform to set them back to the original data. Each of those actions involves navigating a menu. If you try to do this manually, you’ll see why I started wondering about duplication.

Longer Reuse Walkthrough With Screenshots

Until you’re very familiar with the syntax of operations or have some to reuse and update, the first step will be creating a project in OpenRefine. Operations are performed based on column name, so you will need to be sure that the column names in this data source are the same as the ones you plan to reuse it on.

Next, perform the series of actions you plan to repeat on other projects. While you can also perform actions you don’t plan to export and reuse, the operations shouldn’t depend on anything you’re not exporting. If you need to perform a manual action in the middle, like something which needs visual review, you’ll need to export the sets of steps on either side of it into two phases.

Once you’ve completed the actions you wish to repeat, click the “Undo/Redo” subsection of the left sidebar.

Click “Extract.”

In the window that pops up (modal), select which functions you want to export. The preview on the side will update with the relevant code. Use the Export button to generate a text file and save it somewhere appropriate.¹

The easiest way to test how these functions work is to simply use the Undo to bring your project back to the last step before the sequence you exported.

Now click the “Apply” button in that same area.

That opens a popup where you can either upload the same text file you exported or paste raw JSON.

Load or paste to see the operations:

Click “Run operations”

Voila!

As long as the other projects for which you’re using it have the same column names and data that can be parsed the same way by these operations, you should be able to save time in the future. In my case, I am running the same report against different sets of record IDs in the system and getting an item-focused export, so my data should always look as consistent as any data from the ILS does².

Much of the time, my OpenRefine projects are different enough that knowing this wouldn’t have been useful. But with an ILS migration on the horizon, I suspect I’m going to be doing many OR-based analyeses of different sets of records for the same things. I’m hoping this can make it a little less tedious.

Sample Data

If you’d like to test on a sample project, I’ve made one for you!

CSV file - some very basic data.
Operations file - a file containing the JSON of operations.

Simply:

import the CSV into OpenRefine with its standard defaults,
go to the Undo/Redo, where you’ll see “0. Create Project”,
click Apply,
Browse and upload the history.json file,
Run Operations

and you should see something that looks like this:

You could also copy the text and save it somewhere, like Joplin, if that fits your workflows better. ↩︎
This is a joke. But in this case, the operations are consistent and simple enough that even an 856 with no subfield coding and no indicators won’t cause a problem here. ↩︎

Cynthia Ng: Reflection: My eighth year at GitLab and working on Product Operations reporting to the Chief of Staff to CPMO

2026-04-22T00:25:28+00:00

It’s about 2 months early for my 8 year work anniversary reflection. However, my work situation has changed, and it feels like the right time to be reflecting on the past (almost a) year. It’s interesting that 8 years sounds like a long time, but since I’ve changed divisions a couple of times, it also … Continue reading

Web Science and Digital Libraries (WS-DL) Group at Old Dominion University: 2026-02-03: Detecting and reconstructing trustworthy edit histories using web archives

2026-04-21T17:23:38+00:00

The CDC ACIP vaccine recommendations page includes a human-readable (left) and machine-readable (right) last updated date.

In early 2025, US federal websites were undergoing rapid changes with the new presidential administration. In February 2025, we analyzed the rate of page deletions across administrations, and further analyzed the content of changed CDC pages in summer 2025. One surprising thing that we found was that while CDC pages contained a “last updated” date, many of the dates were incorrect: pages contained silent, unannounced updates, generally related to relevant executive orders.

How does trustworthiness change over time?

Silent, unannounced changes that do not match the last modified date of a webpage reduce the trustworthiness of the page. We can study three different change markers: the HTTP last-modified property, meta tags that contain last updated information, and text on the page that is viewable to the user containing update information. These markers are listed in order from most machine readable with standard syntax to most human readable with flexible syntax. The CDC webpages do not have the HTTP last-modified property, but have the other two markers. While the server should automatically update the HTTP last-modified property, the other two properties could have varying levels of automation, from being linked to updates in a content management system to being manually updated by a human. A page with content edits that are always reflected in its last updated markers is trustworthy. A page that has content edits that do not match its last updated markers is less trustworthy. Since we can use web archives to find these changes, we can analyze the trustworthiness gaps of the last updated dates, how often the gaps occur, and if that rate changes over time.

In February 2025, we analyzed the changing rate at which webpages were deleted on US government websites. We found that different presidential administrations have completely different webpage deletion rates, with higher deletion rates correlating with Republican presidential administrations since 2008. Similarly, Tsoukaladelis et al. analyzed silent, unannounced changes on news article webpages in 2022 and detected a correlation between the Allsides media bias score (both left and right) and amount of silent changes by publisher.

Future work in this area will (1) determine a baseline for silent changes on government websites by administration, (2) determine baselines for the news publishers identified by Tsoukaladelis et al. to examine how the silent update rate changes over time, and (3) identify a third type of webpage exhibiting this phenomenon to analyze the change over time as well.

What features contribute to trustworthiness, and how can web archives currently be used to further support or refute trustworthiness?

Figure 2: Change Presentation Continuum. CC BY-NC-SA, adapted from Wikimedia Commons

Figure 2 shows our change presentation continuum. Each website can be categorized based on both its initial properties presented on the live web, as well as from additional captures available on web archives. Table 1 shows an example of each type of change on the continuum along with a description of each type of change presentation.

Change presentation (most trustworthy to least)	Explanation	Example
Past versions	All past versions of the webpage are available to view, giving the user the highest level of trust	Wikipedia All past versions are available for every page
Updates summary	The most recent version of the webpage is available to view, along with a dated list of updates	https://web.archive.org/web/20040128055949/http://immortalised.net/lupdates.html List of changes with dates the changes were made
Update summary	The most recent version of the webpage is available to view, along with a dated description of the most recent update	https://www.merriam-webster.com/dictionary/quixotic Last update date and sentence describing change
Update date correct	The webpage contains a date representing the most recent update, but no update summary	https://www.cdc.gov/winter-weather/safety/index.html Last update date but no information on what was updated. A web archive can be used in order to verify the correctness of the date
Update date incorrect	The webpage has been changed more recently than the update date, negatively affecting trustworthiness
Copyright date	The webpage contains a copyright date, which could be used to infer a most recent change date	https://www.fairfaxcounty.gov/topics/copyright-privacy The only date on the webpage is the copyright date, which is different from the current year, inferring no changes since the new year.
No date	There is no date information anywhere on the webpage, giving the user no information about any changes that have occurred	https://info.cern.ch/hypertext/WWW/TheProject.html There is no date information on the webpage.

Table 1: Change presentation continuum examples based on live web presentation

As shown in the two examples below, web archives can be used to either improve or deteriorate a page’s trustworthiness rating. This means the rating of any site can change more towards either of the extreme ends of the continuum, by using web archives to provide additional captures.

Example 1: more trustworthy: Rakuten Viber Messenger’s Terms of Service includes both a last updated date and a summary for the most recent update. Based on these live web characteristics, it would be labeled “Update summary” on the continuum. The Wayback Machine contains a few captures each month of this webpage. After examining captures in 2025, we can conclude this page is updated a few times a year. Using the additional information from the captures in web archives, we could increase the trustworthiness level from “Update summary” to “Updates summary” for this webpage.

The current change presentation wording (update summary level) is “Last updated: October 21, 2025. We’ve recently updated our terms and policies. View the summary of changes here.” Using web archives, this could be expanded to include more change information, specifically an updates summary list, which is a higher level of trustworthiness.

Updated October 21, 2025 (diff). View the summary of changes here.
Updated May 22, 2025 (diff). View the summary of changes here.
Updated March 24, 2025 (diff). View the summary of changes here.

Example 2: less trustworthy: CDC webpages contain a last updated date, as shown in Figure 1. The initial level for these webpages based on the live web would be “Updated date correct.” In our work, "Coming Back Differently: An Exploratory Case Study of Near Death Experiences of Webpages," we showed that the last updated dates on CDC webpages were inaccurate: words on the pages not in compliance with early 2025 executive orders were removed without updating the last updated date. Therefore, by using web archives, the trustworthiness of these webpages has decreased to the “Updated date incorrect” category.

Figure 3: This CDC webpage experienced changes between January 24 and February 10, as shown in the Wayback Machine Changes Tool, but the last updated meta tag date for both pages is January 8. From Frew et al., Coming Back Differently, Figure 4.

How could web archives be further used to detect and reconstruct trustworthy edit histories?

In the continuum shown in Figure 2, the most trustworthy level is categorized as past versions, and an example website meeting this level is Wikipedia. In order to guide our work of using web archives to amplify the trustworthiness (or lack thereof) of a page’s change presentation, we surveyed the features of the edit histories shown to users on Wikipedia and how researchers used those features to guide their work. We examined peer-reviewed publications from 2018 - 2025 that contained the phrase “wikipedia edit history.”

Authors: Researchers used authorship information about which articles the same author edited, parse edit conflicts on a page, count edits, track the frequency of the edits of an author over time, verify the trustworthiness of individual authors, and also used IP addresses as a proxy for author data of anonymous editors.
Edit properties: Researchers used page edit counts, the “minor” flag, and the time of the edit in their work.
Filtering: Researchers filtered the data (every edit on Wikipedia since inception) by single article, article subject, time of edit, and tag.
View: Researchers viewed the data as a graph, tuple, time series, or as text (for natural language processing or large language model training).
Researchers also used the edit history data to follow redirects and to identify vandalism.
Data: about half of researchers used a derived, cleaned data set and the other half used either Wikipedia dumps or other raw downloads.

Clearly, Wikipedia edit history is extremely useful to researchers who are looking for examples of a variety of edit types. So, why would researchers have a need for web archives when so much Wikipedia edit history data is already available to them? The answer is, not every research would need web archives, but some would. It depends on what type of changes the researcher needs examples of for their work. There are three pathways: cases where Wikipedia edit history contains information not found in web archives, such as when author information is needed; cases where either Wikipedia or a web archive could suit the needs of the researcher, in which case the cleaned and semantic data from Wikipedia would probably be more suitable; and finally, cases where the data in a web archive would be more suitable than Wikipedia. Faruqui et al., and others, have shown that the language on Wikipedia is different than in other contexts, so this is a good starting point for coming up with additional web archive-preferred uses.

Conclusion

We found that websites are communicating inaccurate last updated dates, which affects their trustworthiness. We and other researchers have found that the change rates of silent updates changes over time. We enumerated levels across a change presentation continuum on the live web, and showed how web archives can be used to provide further evidence for or against a webpage’s trustworthiness in this manner. We conducted a literature review of Wikipedia edit history use cases, and used that to start informing how web archives can be used to detect and reconstruct edit histories in a way that will be useful to researchers.

References

Frew et al. Establishing a Baseline by Administration for the Takedown of US Government Webpages using Web Archives.
Frew et al. Coming Back Differently: An Exploratory Case Study of Near Death Experiences of Webpages.
Tsoukaladelis et al. The Times They Are A-Changin’: Characterizing Post-Publication Changes to Online News.
Faruqui et al. WikiAtomicEdits: A Multilingual Corpus of Wikipedia Edits for Modeling Language and Discourse.

-Lesley

Web Science and Digital Libraries (WS-DL) Group at Old Dominion University: 2026-04-21: Was Snopes.com making silent updates to its articles before 2021? [Rating: TRUE]

2026-04-21T17:23:11+00:00

Figure 1: Snopes.com has no public captures on the Internet Archive’s Wayback Machine prior to October 2021.

Snopes.com is a well-known fact checking website. It has been rated trustworthy by multiple rating organizations [1]. However, in 2021, BuzzFeed broke a story that, in fact, Snopes was plagued by plagiarized articles, and that authors were being told to plagiarize and then silently make updates to their articles after publication in the name of SEO.

One of the byproducts of the BuzzFeed investigation is that there are no publicly available captures of Snopes.com in the Wayback Machine prior to 2021. Snopes’s official statement regarding this is that the founder, who was named responsible for the plagiarism, had a policy against Wayback captures; but, now that he was removed from duty, the board was going to change the policy going forwards. In fact, having no captures before 2021 means that the archives available give a more trustworthy historical view of Snopes.com than exists in reality. This kind of manipulation is a form of data craft that misuses web archives [2]. Regardless, users do not have a way to access snopes.com captures before 2021 in the Wayback Machine.

Finding archived copies of Snopes.com before 2021

The Wayback Machine is not the only web archive. In addition, the Wayback Machine accepts donated crawls from other organizations. One of these organizations is Common Crawl, which started crawling in 2008. It seems almost certain that Common Crawl would include snopes.com captures, though probably not with enough frequency to show the silent edits on individual article pages. Searching in a random month (July 2016) within the plagiarism window shows that there are over 12,000 captures of snopes.com pages and resources during that month. Only 453 of these pages were archived a single time during this month, so future analysis with this approach may be possible.

Figure 2: In July 2016, there were over 12,000 captures of snopes.com pages via Common Crawl, with most having 2 or more captures. Source: index.commoncrawl.org

Another organization that has captures of snopes.com before 2021 is Archive-It. Specifically, one of the very first public captures of Snopes.com in the Wayback Machine belongs to an Archive-It organization which turns out to be Mark Graham, the Director of the Wayback Machine. Mark’s personal collections have been instrumental in analyses about news websites [3] and websites in other countries [4]. He has over 300 Snopes articles archived with frequency starting around 2018 in his collection.

We also used MemGator to investigate if there are pre-2021 captures of snopes.com in additional archives; we found that there are over 13,000 captures of the main page pre-2021 in archives including the Icelandic Web Archive, the Australian Web Archive, Archive Today, the Portuguese Web Archive, the Government of Canada Web Archive, along with Archive-It. Over 12,000 of the captures are on Archive-It.

Looking for Silent Updates

Previously, we found silent updates occurring on federal websites in February 2025 [5]. Other researchers also found silent updates on news websites in 2022 correlated with political bias of the news organization [6]. We are also researching how websites announce their changes and how trustworthy the announcements are [7]. Since Snopes.com is regarded as a trustworthy website, we would not have expected silent updates prior to reading the BuzzFeed article.

First, let’s take a moment to observe the current change presentation of Snopes.com. Snopes.com includes a last updated date, and a list of all changes. This would mean that the live web version of Snopes is presenting itself as “Updates summary,” the second most trustworthy level, according to the change presentation gradient [7].

Figure 3: Example of updates summary presentation on a snopes.com article.

However, using Mark Graham’s captures, we have found a number of pages with silent updates.

Example 1: Missing last updated information

The news story “White House Press Secretary Blasted for Sharing Infowars Video to Bar Reporter” states it was published on November 8, 2018. However, a quote from the article states, “On 11 November 2018, White House Counselor Kellyanne Conway admitted…” Based on the published date of November 8, this doesn’t seem possible. Examining a capture from Archive-It, this article used to include a manual last-updated date of November 12. It doesn’t appear Snopes included an updates summary in 2018 like it does in 2026. Sometimes, examining the machine-readable headers gives additional information; in this case, however, the x-archive-orig-last-modified header is the same as the capture time. The Snopes sitemap for that month indicates the page has been updated as recently as March 2025.

Figure 4: This snopes.com article originally stated it was updated, as shown above in an Archive-It capture. On the live web, as shown below, there is no mention of an update on this article anymore.

Example 2: Removal of AP articles

Snopes.com used to include exact text versions of Associated Press articles. Now, however, they redirect to the actual AP article. Given the plagiarism backhistory, this is not surprising. A capture on Archive-It shows the direct text of an AP article on snopes.com. The article link on snopes.com today redirects to the AP article rather than duplicating the text.

Example 3: Silent Update - Word-switching

The most interesting silent update we found was an article where many words had been replaced with synonyms after the fact. The article, “Did Obama Admin Build Cages That House Immigrant Children at U.S.-Mexico Border?” was published July 2, 2019 and includes no update date. However, we examined the capture from July 30, and it contained completely different text. Searching the web for this original text shows multiple citations. However, by August 9, the text changes to the version that matches the live web.

Figure 5: Differences between the July 2, 2019 and live web snopes.com article “Did Obama Admin Build Cages That House Immigrant Children at U.S.-Mexico Border?” show that the article was edited with a word-switching pattern.

Example 4: Updates declaring SEO

David Mikkelson’s 2007 article on Mr. Rogers had an SEO update in 2022, which was disclosed in its updates summary. The original title is widely referenced, as determined by searching the web. If you search the web for “Updated SEO” on snopes.com, you can find other articles where they have noted updating the title as well. One of the reasons why Snopes has been rated more trustworthy and less biased by the organization Media Bias / Fact Check is because its titles are questions, which is interpreted as more neutral than a title that supports one conclusion over another.

Figure 6: Google search for “Updated SEO” on snopes.com showing multiple articles with titles changed.

Outlook

Originally, we thought that a trustworthy website like Snopes.com would not have silent updates. However, after reading the BuzzFeed article about the rampant plagiarism on the site, we decided to investigate. We found evidence of silent updates using web archives beyond the Wayback Machine. This reduces Snopes.com’s trustworthiness level. Future work could determine if their plagiarism problem has truly ended, or if silent updates are still occurring post-2021.

References

[1] Yang et al. Are Fact-Checking Tools Reliable? An Evaluation of Google Fact Check. https://arxiv.org/abs/2402.13244v1

[2] Acker et al. The weaponization of web archives: Data craft and COVID-19 publics. https://misinforeview.hks.harvard.edu/article/the-weaponization-of-web-archives-data-craft-and-covid-19-publics/

[3] Weigle et al. Right HTML, Wrong JSON: Challenges in Replaying Archived Webpages Built with Client-Side Rendering. https://doi.org/10.1109/JCDL57899.2023.00022

[4] Ben-David et al. The Internet Archive and the socio-technical construction of historical facts. https://doi.org/10.1080/24701475.2018.1455412

[5] Frew et al. Coming Back Differently: An Exploratory Case Study of Near Death Experiences of Webpages. https://digitalcommons.odu.edu/computerscience_fac_pubs/404/

[6] Tsoukaladelis et al. The Times They Are A-Changin’: Characterizing Post-Publication Changes to Online News. https://doi.org/10.1109/SP54263.2024.00033

[7] Lesley Frew. Detecting and reconstructing trustworthy edit histories using web archives. https://ws-dl.blogspot.com/2026/02/2026-02-03-detecting-and-reconstructing.html

--Lesley

Meredith Farkas: Neoliberal Time and the Promise of Slow Librarianship

2026-04-20T20:34:06+00:00

I’ve meant to post about this for a while, but one of the book chapters I wrote back in 2023-24 during my sabbatical has finally come out (the other is estimated for early 2027, <sigh…>). The book Slow Librarianship: Reflections and Practices edited by the wonderful Ashley Rosener has finally been published by Library Juice Press! I deposited a copy of my chapter, “Neoliberal Time and the Promise of Slow Librarianship” in Knowledge Works Commons so that anyone can freely access it. I’d love to hear your thoughts about it! I know a lot of people writing about slow librarianship have cited my blog posts because there isn’t a lot out there on slow librarianship and I’m not a big scholarly publishing type, so hopefully this will be useful to those folks looking for a more traditionally-published source to cite.

I’m also really honored to be keynoting the Conference on Academic Library Management (CALM) next month, talking about slow management in a fast world. Ironically, I had thought of submitting a proposal to give a talk on slow librarianship at the conference but talked myself out of it because I’m not currently a manager. It’s amazing that after 20 years in this profession, I still suffer from impostor syndrome. CALM has consistently been my favorite library conference to attend and it’s a conference that has embraced slow practices in everything they do so I’m especially honored to have been asked to speak. The conference is free to attend and my talk will be recorded, so I’ll be sure to share it once it’s up on the web.

Dan Cohen: The Role of a New Machine

2026-04-20T18:08:51+00:00

Marcin Wichary, Data General keyboard, CC BY-NC 4.0

Stop me if you’ve heard this one before:

A crack team of hardware and software engineers, inspired by breakthroughs in computer science and electrical engineering, are driven to work 18-hour days, seven days a week, on a revolutionary new system. The system’s capabilities and speed will usher in a new era, one that will bring transformative computing to every workplace. The long hours are necessary: the team knows that every major computer company sees what they see on the horizon, and they too are working around the clock to take advantage of powerful new chips and innovative information architectures.

The team is almost entirely men, men whose affect and social skills cluster in a rather narrow band, although they are led by a charismatic figure who knows how to persuade both computer engineers and capitalists. This is a helpful skill. Money, big money, is flowing into the sector; soon it will overflow. Engineers are constantly poached by rival companies. Hundreds of new competitors arise to build variations on the same system, or to write software or build hardware that can take advantage of this next wave of computing power. Some just want to repackage what the computer vendors produce, or act as consultants to the companies that adopt these new machines.

The team solves one problem after the next, day and night, until the machine is complete. They focus, overfocus, block out the other concerns of the world. Their wives are ignored, as are the kids. The work is too important.

* * *

Such is the story of Data General and the group that built the computer system code-named “Eagle,” which would be successfully marketed as the Eclipse MV/8000. My summary above comes from Tracy Kidder's wonderful book The Soul of a New Machine, published in 1981. It’s about the rise of minicomputers, a now-amusing name for machines the size of double-wide refrigerators, which were considered a major advance during the 1970s, when gargantuan IBM mainframes still roamed the earth and were possessed only by the largest companies and bureaucracies. Minicomputers used new CPUs and memory that made computing accessible to a much wider range of applications and locations, and were relatively cheap. They flourished.

The Soul of a New Machine has much to recommend it — it won the Pulitzer Prize for its propulsive narrative and crisp explanations of complex technology — but I’m writing about it now, following Kidder’s recent passing, because the book helpfully dislodges you from your presentist perspective and asks, “Look what happened before — sound familiar?”

* * *

A half-century after it was published, The Soul of a New Machine does a better job challenging AI hype than most current criticism. (Also, there are probably writers working on books about AI who are shaking their fists at Kidder for beating them to that memorable title.) It's hard to read The Soul of a New Machine in 2026 without wondering whether all this AI hype is really so new. Is AI truly more revolutionary than a previous wave of computer technology that offered, for the first time, to put screens on every desk of every company? The Data General team helped to bring about a transition not from existing software and hardware to incredibly intelligent software and hardware, or from powerful computers to superpowerful computers, but literally from paper to digital files and high-speed processing. Now that is a transition. The millions of companies that could not afford an IBM mainframe could afford a Data General Eclipse or a DEC VAX system or a minicomputer from another competitor. They could, for the first time, give every employee the power of computers. Is having Microsoft Copilot help your accountants with their spreadsheets more revolutionary than moving those accountants from physical spreadsheets to electronic ones?

Amazingly, the final chapters of The Soul of a New Machine tackle exactly the same profound questions we are struggling with today regarding the impact of artificial intelligence, and Kidder records Data General engineers expressing concerns that sound straight out of the mouths of engineers working at OpenAI or Anthropic. The team’s excitement upon the completion of the Eagle leads to reflections and bigger worries than beating their competitors. What if the Pentagon wants to use the Eagle for war or other destructive purposes? Should the team object or build back doors into the machine? What will the new computer system mean for employment, since it will replace many functions of work with software, and do those tasks faster than anyone can imagine, in nanoseconds? What if their work culminates in true artificial intelligence, and the machines take over and destroy us?

Spending so much time with the team, Kidder begins to ponder these questions himself — and has his own unsettling encounter with the technology. An engineer introduces Kidder to Adventure, one of the engrossing text games of early computing. He is sucked into its digital world, playing nonstop for hours. The computer suddenly feels alive, intelligent.

But Kidder pulls back.

It was the time of night when the odd feeling of not being quite in focus comes and goes, and all things are mysterious. I resisted this feeling. It seemed worth remembering that Adventure is just a program, a series of step-by-step commands stored in electrical code inside the computer.
How can the machine perform its tricks? The general answer lies in the fact that computers can follow conditional instructions.

Kidder turns to one of the Data General engineers, Carl Alsing:

I asked Alsing how he felt about the question — twenty years old now and really unresolved — of whether or not it's theoretically possible to imbue a computer with intelligence — to create in a machine, as they say, artificial intelligence.
Alsing stepped around the question. “Artificial intelligence takes you away from your own trip. What you want to do is look at the wheels of the machine and if you like them, have fun.”

Alsing’s focus on the role of the new machine in your life or work, rather than its purported soul, instantly dispels the mythology surrounding this emerging technology. Even after an all-nighter building a revolutionary computer, Alsing is lucid about what he is making: a tool that might be helpful for some people and some purposes, but not for others.

* * *

In the 1980s, most of the minicomputer companies, launched with such excitement in the late 1970s, failed. Data General was acquired for a fraction of the billions it was once worth. The minicomputer, however, was broadly adopted, was transformative, became routine, and then was surpassed by a new new machine, the personal computer.

Later, Data General’s domain name, DG.com, was sold to a chain of discount stores, Dollar General.

David Rosenthal: Angels in America

2026-04-20T02:29:36+00:00

I have wanted to write this post for a long time, but I was waiting until I could visit the invaluable Royal National Theatre Archive to check my memory of their early productions. It doesn't look like I'll be in London any time soon, and I have the time now to write a long post about a long play, so here goes.

Growing up in London meant that theatre has always been an important part of my life. I have seen a great many plays including some legendary performances and magnificent productions, such as Royal National Theatre's 2014 King Lear. One of my particular theatrical interests is long-form plays. Highlights of this genre have included:

Play Text

Ken Campbell's Science Fiction Theatre of Liverpool's 9-hour Illuminatus Trilogy
The Magic Theatre's all-night production of Murray Mednick's Coyote Cycle
The Royal National Theatre's 6-hour adaptation of Phillip Pullman's His Dark Materials
Taylor Mac's 3x8-hour A 24-Decade History of Popular Music

But there is one such play that is very special to me, Tony Kushner's 7+ hour Angels in America. It is clearly among the greatest plays of the 20^th century. I was there at the beginning, and I have seen many productions since. Below the fold I recount my history with this masterpiece.

Introduction

Anyone interested in this play should read both the text of the two halves, Millennium Approaches and Perestroika, and Isaac Butler and Dan Kois' magisterial and comprehensive oral history, The World Only Spins Forward: The Ascent of Angels in America. Because my story starts in 1991 I have used both to refresh my memory. Below the many quotes without links are from Butler and Kois, to whom I owe a debt of gratitude. I also viewed the National Theatre's 2017 production on the National Theatre at Home streaming service.

When I moved to the Bay Area in 1985 It was a decade since I'd lived in London and I was starved of theater. So I went a bit nuts and over the next few years subscribed to American Conservatory Theater, Berkeley Repertory Theatre, the Magic Theater and the Eureka Theater.

Eureka Theatre (1991)

The story of the play starts with a $50,000 grant from the National Endowment for the Arts for Tony Kushner to write a "two-hour play, with songs" for "five gay men and an angel" that the Eureka would produce. In 1989 the play was developed and in 1990 workshopped at the Mark Taper in Los Angeles.

KUSHNER: I wrote the part of Harper for Lorri Holt, Hannah for Abigail Van Alyn, Sigrid [Wurmschmidt] was the Angel. And Jeff King, I wrote the part of Joe for him. And that took care of the Eureka company. My first year at NYU, I became friends with Stephen Spinella. I thought then, as I think now, that he was one of the most remarkable actors I'd ever met, and I loved writing for him, and so I wrote Prior Walter for him.

As a subscriber to the Eureka I had responded to their call for donations to stage Angels in America in their next season, so I was anxious to see it. By the time it arrived at the Eureka it had evolved into two long plays with five gay men, two women, an angel and no songs.

I believe I saw Millennium Approaches the weekend after it opened, and Perestroika the following weekend. The cast was different from that at the Mark Taper. Rick Frank (Roy) and Sigrid Wurmschmidt (Angel) had both died, and Lori Holt had a new baby. It was:

Hannah: Kathleen Chalfant
Roy: John Bellucci
Joe: Michael Scott Ryan
Harper: Anne Darragh
Belize: Harry Waters Jr.
Louis: Michael Ornstein
Prior: Stephen Spinella
Angel: Ellen McLaughlin

The Eureka was staging Millennium Approaches, a four-hour play full of scene changes and magic, with almost no money. So another abiding memory is that they got this enormous impact with an incredibly stripped-down production:

[Ellen] McLAUGHLIN: Not that many people saw the Eureka version of it, but it was very important to those who did. I think there was a kind of beauty to the hammer and nails and spit and Scotch tape quality of that first version. It was moving because we had nothing.

In some ways it reminded me of the San Francisco Mime Troupe's annual free shows in parks around the Bay Area. The same quality of conspiring with the audience's imagination:

KATHLEEN CHALFANT: It was in some ways the most beautiful version of the play, and the most Poor Theater version of the play.
[Dennis] HARVEY: They basically had a giant shower curtain in front of the stage. For scene transitions they would just whip the shower curtain across, one actor at the front and one at the back, and when they got the other side it would be a new scene.
KUSHNER: To this day no one has ever done better with the magic. David [Esbjornson] is incredibly clever designing and building gizmos, so every magic trick in the play, David figured out a way to do it. There was no money or anything. He built all this shit — it was incredible.
DEBORAH PEIFER: That sense of amazement of a book popping up out of the floor in flames, all done with lighting.
KUSHNER: He did it all with bungee cords.

My most abiding memory of that first part was walking out of the theater to my car after midnight realizing I had seen the birth of a masterpiece. Theater critic Deborah Peifer sums up my reaction:

PEIFER: I have never in my life seen a situation in which people did not leave the theater during the intermission unless they had to. And I'm not talking about Can I get a cup of coffee? but Can I make it through the next act without a bathroom break? People could not bear to be out of that theater while this thing was happening.
To call this a brilliantly realized, profoundly funny, wickedly thoughtful piece of theater is to discover the severe limitations of language. I find myself wanting to say simply, it's more than I ever imagined. This is an experience in the theater you will remember for your whole life.

Deborah Peifer, Bay Area Reporter, May 30 1991

Perestroika was even more stripped-down, little more than a staged reading:

KUSHNER: Originally, every act of the five acts of Perestroika started with a clown scene set in the Soviet Union. These ended up being the first five scenes of my play Slavs! [1994].
ESBJORNSON: I used the five Bolsheviks as curtain raisers. I made the actors hold the scripts in hand while they moved around. And then at one point in each act, they laid down their scrips and acted out what I considered to be the central point of that act.

It wasn't just that there were five acts, but each of them was rather long. Butler and Kois' description of the first night matches my later recollection of how long it was:

[Brian] THORSTENSON: It got to the scene between Hannah and Prior where Prior's in the hospital and Prior says "I've always depended on the kindness of strangers." They finished the scene and the audience erupted into this ... applause ... I think it lasted a good five minutes. Kathleen and Stephen looked out at the audience, like, What is going on?.
McLAUGHLIN: I came out late into the evening as the Angel wearing the wings and the whole get-up, stood in front of the curtain and said, Act 5: Heaven, I'm in Heaven.
And the woman in the front row said "Act FIVE?! Oh my GOD! DO YOU KNOW WHAT TIME IT IS?!"
And I said "No". Because I honestly had no idea. It's not like I was wearing a watch.
And she said "It's MIDNIGHT, for God's sake! What's going on with the playwright? ACT FIVE? How long is it?
And I said, :We've never done it so I don't know, maybe forty-five minutes?" And she said, "The buses aren't even running anymore! How are we supposed to get HOME?" And she turns to the rest of the audience and says, "Are we going to stay?" And people sort of nodded and mumbled and she says "Well, I guess we'll stay, but I mean really ..."
And then she said, "But that's the end, right? There isn't an Act 6 or something?"
And I said, "Well, there's an epilogue."
And she said, "Oh my GOD, is he NUTS? An EPILOGUE? How long is THAT?"
And I said, "Well, apparently we HAVE TO STAY, but this is RIDICULOUS. TELL HIM HE HAS TO CUT!"
And then I said "Well, the longer we keep talking here ..."

Millennium Approaches was a real play and, despite being over four hours, had the audience in the palm of its hand with rapt attention. Perestroika was really different. Because it was clearly a work-in-progress, the audience felt that they were part of the process of creation, willing the show into existence.

Sometimes at the Berkeley Rep's Ground Floor residency program for new work the teams show their work — an example was Julia Cho's Aubergine which I saw both as a work-in-progress at the Ground Floor and next year in the Rep's season. Even as works-in-progress these shows are way shorter and way more polished than this Perestroika, and there is none of that show's unique, intense audience involvement. Of course, as the Angel notes, this was heightened by the show's length:

McLAUGHLIN: And then after the show, as the actors were basically limping to the dressing rooms, Tony, looking sort of glassy-eyed, came over to us and said, "You know, a really interesting thing happens after and audience has been in the theater for a really long time, they start to lose their bearings and become very malleable. They, like, forget what the think they believe about things and what they do for a living and their names and where they live and ..."
And we were like, "Yeah, Tony, and you really have to cut it."

It was magnificent but it killed its host. Butler and Kois quote the Eureka's business manager:

ANDY HOLTZ: That was the end of the Eureka Theatre as a producing company, The play that cemented the Eureka's place in the history of American theater was also the play that was too epic for such a small company. It's, like, the mom died giving birth to this amazing baby.

Royal National Theatre (1992)

Perhaps the most astonishing thing in the play's whole history is that, apart from a workshop at Juillard, the next production of Millennium Approaches was at the National Theatre in London. At the time, the National Theatre's productions on their two big stages, the Olivier and the Lyttleton, were pretty conservative, as befits the national flagship. But they also had the Cottesloe (now the Dorfman). It is essentially an empty cube, with tiers of seats on two sides. It can be configured in many different ways. For example, for Sing Yer Heart Out For The Lads most of the floor was arranged with tables and chairs, with the audience there being some of the patrons of the pub.

The National Theatre has a history of more adventurous productions in the Cottesloe; it opened with Ken Campbell's Science Fiction Theatre of Liverpool's Illuminatus Trilogy featuring drugs, satanic rituals, blasphemy and nudity. The trilogy later moved to The Roundhouse, which is where I saw this marathon. My main memory was that between the plays meals were served in the lobby. The actors ate with the audience, staying in character.

Nevertheless, Richard Eyre, the artistic director, took a huge risk:

RICHARD EYRE: Gordon Davidson sent me the play and said,, "I think you'd be interested in this". By page 2, I'd decided I wanted to do it.

He chose Declan Donnellan of the Cheek by Jowl theatre company to direct it, and Nick Ormerod, Donellan's partner, to design it. I'd seen several Cheek by Jowl productions at the National Theatre. They did classical plays, so Kushner took them to New York:

DONELLAN: Sometimes when you see images of New York, you think Oh, it's not authentic New York. It's performed New York, from movies and television. But when you get to New York, you find that New York is performing itself. Everybody is ready for their close-up.
ORMEROD: In delis and diners and whatever, they act like New Yorkers they've seen in the movies.

The cast was:

Hannah: Rosenmary Martin
Roy: Henry Goodman
Joe: Nick Reing
Harper: Felicity Montague
Belize: Joseph Mydell
Louis: Marcus D'Amico
Prior: Sean Chapman
Angel: Nancy Crane

NT's 1993 Angel

David Milling was the stage manager:

DAVID MILLING: The staging was incredibly simple. It was a shiny black floor and a giant American flag as the backdrop. And then in the center of the flag there were small doors for pieces of scenery to run through. Only at the end of the play did the flag split, half going left, half going right, and the Angel tracked through in a cloud of smoke.

I'm sure that the first thing everyone who saw the show remembers is the shock at the end of the Angel bursting through the flag with a huge noise, lots of smoke and a blinding light then announcing:

ANGEL: Greetings, Prophet;

The Great Work begins;

The Messenger has arrived.

(Blackout.)

But the start was almost equally memorable:

JON MATTHEWS: It opened with this image, there was nothing on the stage, and the furniture is on the sides, and they're sitting along the sides, and there was this balloon globe, and it had this light inside it, and they all put their hands on it, and then the play began.

Donellan said "My production was very much about the maintenance of tension", and I remember the production as a headlong charge forward:

KUSHNER: Caryl Churchill saw one of the early performances and came up to Declan afterwards and said, "Well congratulations, you've solved the short, choppy scene problem." When you do a play with short scenes, the scene ends, the audience has to disengage from where they've just been, and open themselves up to the next thing. That's hard to do because it involves stopping and starting over and over again. What Declan did is he dovetailed the ends of almost every scene in Millennium. He took the penultimate and the ultimate line, separated them, took the first line of the next scene and put it between the two. So you'd already be in the next scene. He wove them all together.

Donellan could do this because the staging was so sparse that it needed no time for scene changes. The actors carried in whatever props were needed for the next scene, and carried off those from the preceding scene.

It is important to understand both the risk the National Theatre was taking, as an institution supported by the government, and why it was so important, especially to the theatre community:

GARSIDE: The politics of it hit on the right moment. We were having our side of the conservative 1980s with Thatcher and the special relationship with Reagan. There was a kind of resentment of America, a dislike of their politics and how it intersected with our politics. And then there was an audience who hadn't seen a play about gay men and AIDS on a large scale, for whom the play was a revealation.

The big legal fight in gay rights at the time was against someting called Section 28. This was the big thing. It was in effect between 1988 and 2003, and barred the "promotion" of homosexuality.

Royal National Theatre (1993)

The next year both parts opened on Broadway and the National Theatre revived Millennium Approaches and added Perestroika in repertory. For the first time, I saw both parts in one day.

MYDELL: So we opened at the National, and you could see Part 1 and Part 2 in one day. That was seven and a half. People did it! We did it, and people came to see it! It didn't seem like — it felt like it was an event more than a play.

The cast was:

Louis: Jason Isaacs
Belize: Joseph Mydell
Angel: Nancy Crane
Joe: Daniel Craig
Hannah: Susan Engel
Harper: Clare Holman
Prior: Stephen Dillane
Roy: David Schofield

Part 1 was familiar, but it was the first time I'd seen Part 2 staged. First, seeing them as a seven and a half hour marathon was a revelation. Millennium ends with the mother of all cliff-hangers as the Angel arrives. Resuming the story after a quick meal is completely different from resuming it a week later. Second, Perestroika was very different from my memory of the Eureka. Kushner had done massive rewrites after the Eureka and the 1992 workshop at the Taper in LA:

KUSHNER: I know I haven't got it right yet. I'm not saying I don't think it's good — I think it's always been a good play, Perestroika — but it's never been a finished play and it never ever will be completely finished.

Many people compare the two parts and rate Perestroika as inferior, citing that its a lot more difficult and the fact that Kushner keeps changing it. But this is likely because they have seen it as two separate plays, which is a mistake. I'm pretty sure that people like me who have seen in in a marathon see it as a single play that changes once the Angel arrives. Change is one of its major themes, after all. And it is very Kushner-esque to have the Angel, whose message is to stop change, be the cause of change in the structure of the play as she is in Prior.

Next time I'm in London I plan to visit the Archive and expand these two sections.

American Conservatory Theater, San Francisco (1994)

ACT Program

I saw ACT's production of both halves, I think on successive weekends, but I remember very little about it. It was directed by Mark Wing-Davey, who played the two-headed Galactic President, Zaphod Beeblebrox, in the radio (my favorite) and TV versions (forget it) of The Hitchhiker's Guide to the Galaxy, written by Douglas Adams. I'd been impressed by his production of Caryl Churchill's Mad Forest at Berkeley Rep.

The cast was:

Hannah: Cristine McMurdo-Wallis
Roy: Peter Zapp
Joe: Steven Culp
Harper: Julia Gibson
Belize: Gregory Wallace
Louis: Ben Shenkman
Prior: Garret Dillahunt
Angel: Lise Bruneau

Dennis Harvey's review noted that:

the director throws action all over the Marines Memorial stage. Kate Edmunds’ set design is dominated by rolling scaffold bridges and graph-patterned backdrops. Their severity suggests a societal infrastructure stripped bare. Huge curtains (one a rather too-obvious American flag), one hydraulic ramp, fully exposed flying rig for the “Angel” (Lise Bruneau), fog, film projection, etc. add to the sensory overload.

This may be one reason it didn't stick in my memory. After stripped-down productions in the Eureka, basically a warehouse, and the National Theatre's Cotttesloe flexible space, the traditional proscenium stage, a more fleshed-out, much flashier staging, and the somewhat distant seating would have been jarring. Indeed, soon after this I stopped subscribing to ACT, only visiting for their excellent productions of Tom Stoppard's plays.

Royal National Theatre (2017)

By dint of waking up very early and standing in line for a long time I got day seats for a marathon of Marianne Elliot's sold-out, extraordinarily impressive production. It was a complete contrast to the earlier version. The cast was:

Hannah: Susan Brown
Roy: Nathan Lane
Joe: Russell Tovey
Harper: Denise Gough
Belize: Nathan Stewart-Jarrett
Louis: James McCardle
Prior: Andrrew Garfield
Angel: Amanda Lawrence

Joe and Hannah

Elliot's staging was a fascinating way to use the National Theatre's huge resources and the Lyttleton's vast proscenium stage to simulate the original's sparse aesthetic. She used multiple revolves and mostly skeletal scenery that flowed in and out to create small patches of light in the darkness to show, for example, the phone call between Joe and Hannah. Occasionally, as for Harper and Mr. Lies in Antarctica, the whole stage was lit but bare. There was only one scene with the kind of lavish scenery one often sees in the Lyttleton. It was the Council Room of the Hall of the Continental Principalities. Kushner's stage directions for this scene fill multiple pages, and the set needs to contrast Heaven with Earth, so this choice made sense.

One of the most striking and memorable things in Elliot's production was her vision for the Angel:

ELLIOT: Every image you see of this play involves a lovely angel in a white dress on a wire. I didn't want that.

Ben Power, the National's deputy artistic director, explains the Angel's entrance:

POWER: Prior's standing on his bed, as in other productions. The lights are changing. The sound of the approaching object is getting louder and louder. It's extremely loud in the auditorium. The lights change around him and he says, "Very Steven Spielberg".

Everyone's eye are on him and they're also going up to the flies. We know what's about to happen. They're going to fly in a woman with wings. As we're looking, as it's all building to a point of climax. At that point of climax there is a sense of a drop and a full blackout, which is very disorienting.

The lights come up. Everyone's eyes are looking up, looking for what object is coming in through the broken roof. Andrew's looking up there. And there's nothing there. As his eyeline comes down, there, strewn on the floor, among the rubble, is this thing. It's a sort of creature mess in browns and blacks. And then it rises from the floor — it's clearly been dropped from a great height — and coalesces into one body.

Lyra and Armored Bears

The National Theatre has resources that few other theaters do. One is a long history and deep expertise in stage puppetry. This reached a peak with His Dark Materials because in the play's world:

humans' souls naturally exist outside of their bodies in the form of sentient "dæmons" in animal form which accompany, aid, and comfort their humans.

Each actor was accompanied by a puppet of their daemon, manipulated by one or more puppeteers in head-to-toe black. It didn't take long for audience members to stop seeing them. At the end of the second part, all 28 actors came out for their curtain call. And then suddenly the puppeteers all pulled off their black head-dress, and you saw there were more of them than there were actors. And then the backdrop vanished and you saw all the way to the rear wall of the enormous Olivier stage. Standing there were all the stagehands. There were more of them than the actors and puppeteers combined. It was an amazing display of the vast resources the National Theatre can command for a major production.

Angel and Prior

Elliot's Angel was accompanied by a set of black-clad "shadows" like the daemon's. Except when Prior and she were wrestling, the Angel wasn't on a wire but being carried by the shadows. They would scurry around on all fours, sometimes converging on her to lift her up or sweep her massive wings, and sometimes heading off to the back of the set.

It wasn't just the physical resources the National Theatre devoted to the production, it was the time:

KUSHNER: I've never seen a director work as long or as hard on a production. A year of preparation. And you can see that degree — the depth of involvement, it's reflected in the design and in many of the choices she's made.
ELLIOT: We spent about a year and a half on the design. Not every day, but we touched in a lot. And I wished I had longer!
...
ELLIOT: We had eleven weeks, longer than anyone else has had.
KUSHNER: In a way it's the first adequate rehearsal period we've had for these plays.

When the production transferred to Broadway, it won the Tony for the Best Revival of a Play, and Andrew Garfield won for Best Actor and Nathan Lane won for Best Featured Actor. Both performances richly deserved the award.

Berkeley Repertory Theatre (2018)

Program

Berkeley Rep's production was directed by Tony Taccone, who co-directed the Mark Taper workshop with Oskar Eustis, and starred Stephen Spinella, for whom Prior was written and who I had seen at the Eureka, as Roy. So I have seen him play both of the victims of AIDS — his portrayal of sickness is remarkable, as was the contrast between how his Prior and his Roy fought the disease.

The cast was:

Hannah: Carmen Roman
Roy: Stephen Spinella
Joe: Danny Binstock
Harper: Bethany Jillard
Belize: Caldwell Tidicue
Louis: Benjamin T. Ismail
Prior: Randy Harrison
Angel: Francesa Faridany. Lisa Ramirez

Again, I saw both parts on a single day, as I recall starting at 1pm and ending at 11pm. It was astonishing how well the Rep's, with a regional theater's resources, stood up to the National Theatre's massively resourced production. This huge play can use huge resources, but it does not need them.

Angel and Hannah

Taccone's Angel was clearly influenced by Elliot's, but lacked the shadows. Despite this the Angel's flying, always the most difficult thing to stage, was really well done.

For the first time I got to see the "Roy in Hell" scene, which almost every production omits. It isn't in the published text. Omitting it means Roy's last appearance is when his ghost encounters Joe, a meeting between the play's two doomed characters. Including it, with Roy bargaining for something to do, is a sort of tribute to his drive and contrasts against Joe's spinelessness.

The Berkeley Rep's program had an interview with Spinella, who was initially reluctant to play Roy:

I got a text from Kushner saying — all I really remember is one word — "vital". That Roy is incredibly vital. I had already gone back and read all the Roy scenes, and it really hit me. That's the fun of playing this guy who is dying. He is fighting it tooth and nail. It's this knockdown, drag-out fight with this person who has this incredible will to live. It's different than Prior, who in a way is running away from his own death. Roy is just trying to get his ducks in a row and he's fighting the disease. He loses constantly, yet he keeps coming back. He is unrelenting, and that appeals to me. I'm not going to be in that hospital bed until I am ready to die. The hospital bed is going to have to grab me and pull me into it.

This could have been a quote from Nathan Lane.

The production gained glowing reviews from, among others, the LA Times and the SF Chronicle. For me, as my sixth viewing, seeing the play come back to the Bay Area over a quarter-century after it started here, over a single day, with such a grown-up staging, was a delight.

Ed Summers: Weekly Bookmarks

2026-04-19T04:00:00+00:00

These are some things I’ve wandered across on the web this week.

🔖 Why I ask AI coders to ask me questions

Next, I will tell the AI to ask me 10 to 20 questions about architectural choices that I haven’t thought about, that it needs to pin down before writing code. Invariably, the questions that I’m asked are detailed, insightful, and pivotal to the design. Sometimes the AI will give me choices for answers, sometimes it will ask the question and make a sceptic recommendation.

I answer the AI’s questions, after which I ask the AI if it has any follow-up questions for me. After a few rounds of this back-and-forth, I ask the AI to create high-level documentation for different aspects of the solution. I then review the documentation. This documentation lets me review the entire design and make corrections, but it also serves as a record that this or another AI will be able to use in the future when it comes back to do software maintenance.

🔖 Gottstein Shoes

Sustainably felt slippers by Gottstein • Made of undyed wool • Comfortable felt slippers • Manufactured in Tyrol, Austria ► Feel the nature on your feet!

🔖 Baabuk Shoes

Baabuk wool footwear made responsibly and sustainably. Designed in Switzerland, crafted in Nepal and Portugal, built for comfort. The most comfortable wool sneakers, wool slippers, and boots.

🔖 Mythos and Cybersecurity

In the short term, we need something simpler: greater transparency and information sharing with the broader community. This doesn’t necessarily mean making powerful models like Claude Mythos widely available. Rather, it means sharing as much data and information as possible, so that we can collectively make informed decisions.

We need globally co-ordinated frameworks for independent auditing, mandatory disclosure of aggregate performance metrics and funded access for academic and civil-society researchers.

This has implications for national security, personal safety and corporate competitiveness. Any technology that can find thousands of exploitable flaws in the systems we all depend on should not be governed solely by the internal judgment of its creators, however well intentioned.

🔖 MNT Pocket Reform

Pocket-sized, repairable, and fully open source: Our iconic MNT Pocket Reform mini laptop is as versatile as you need it to be. Whether you’re traveling, attending classes, programming your tools at the cafe, or working at a data center—Pocket Reform fits nearly any space. And if you want it stationary, hook it up to a big monitor and play a video game or browse the web. Or get creative and use Debian-compatible software such as Libre Office, FreeCAD or Krita to express yourself.

🔖 Tape 05 / Boards of Canada

music video / tune teaser for new album?

🔖 A Philosophy of Software Design / John Ousterhout

Writing computer software is one of the purest creative activities in the history of the human race. Programmers aren’t bound by practical limitations such as the laws of physics; we can create exciting virtual worlds with behaviors that could never exist in the real world. Programming doesn’t require great physical skill or coordination, like ballet or basketball. All programming requires is a creative mind and the ability to organize your thoughts. If you can visualize a system, you can probably implement it in a computer program.

This means that the greatest limitation in writing software is our ability to understand the systems we are creating. As a program evolves and acquires more features, it becomes complicated, with subtle dependencies between its components. Over time, complexity accumulates, and it becomes harder and harder for programmers to keep all of the relevant factors in their minds as they modify the system. This slows down development and leads to bugs, which slow development even more and add to its cost. Complexity increases inevitably over the life of any program. The larger the program, and the more people that work on it, the more difficult it is to manage complexity.

🔖 FAR OFF TOWN - DUNEDIN TO NASHVILLE

Napier film-maker Bridget Sutherland made the 82min doco about one of Dunedin’s favourite musical sons and his trip to Nashville to make the album ‘Frozen Orange’ with the band Lambchop. Kilgour and Lambchop are close friends and have often toured together. They both record for Merge, and the merging of Kilgour with Nashville and its homogeneous Music City product was an ideal subject of intrigue for a film.

But Kilgour did not record with normal mainstream Nashville, he went to the city’s indie underground, and worked with people he knew and liked, and who definitely liked him, like Lambchop’s Mark Nevers, who produced the record in his own studio. Nashville’s front window is a mixture of Old Country - the never-quite-made-its who play on the hour every hour in the bars for tips - and far more lucratively, New Country, which is manicured market-aimed singers who are young, slick, attractive, bland, radio-friendly, and often extremely successful.

🔖 A eulogy for Vim

Vim is important to me. I’m using it to write the words you’re reading right now. In fact, almost every word I have ever committed to posterity, through this blog, in my code, all of the docs I’ve written, emails I’ve sent, and more, almost all of it has passed through Vim.

My relationship with the software is intimate, almost as if it were an extra limb. I don’t think about what I’m doing when I use it. All of Vim’s modes and keybindings are deeply ingrained in my muscle memory. Using it just feels like my thoughts flowing from my head, into my fingers, into a Vim-shaped extension of my body, and out into the world. The unique and profound nature of my relationship with this software is not lost on me.

🔖 easyaligner: Forced Alignment Made Easy

easyaligner is a forced alignment library for aligning text transcripts with audio. It is designed with a focus on ease of use, flexibility, and performance. The library can be used for a variety of applications, including

Aligning e-texts with audiobook recordings to create interactive reading experiences (see the interactive demo below).
Aligning podcast transcripts to enable features like chapter navigation and keyword search.
Aligning protocols and recordings of parliamentary debates for research and accessibility purposes.
Fixing misaligned subtitles in videos, or creating new subtitles from transcripts.
Creating large-scale speech recognition and speech synthesis datasets for AI model training.

🔖 The Midnight Sky

The Midnight Sky is a 2020 American science fiction film directed by George Clooney based on the 2016 novel Good Morning, Midnight by Lily Brooks-Dalton. The script was written by Mark L. Smith. Clooney plays a leading role in his film, as an aging scientist who must venture across the frigid Arctic Circle to warn off a returning interplanetary spaceship following a global catastrophe on Earth. Felicity Jones, David Oyelowo, Tiffany Boone, Demián Bichir, Kyle Chandler, and Caoilinn Springall also star in this film.

🔖 Quo Vadis, Crawlers? Progress and what’s next on safeguarding our infrastructure

Readers, contributors, responsible bots, and abusive bots all share the same access points to our websites and infrastructure. We have therefore orchestrated our work with maximum care to minimize impact on our reading and editing community, with the ultimate goal of not impeding any person from accessing our projects. As a result of this work, we’re currently blocking or throttling about 25% of all automated requests that are coming from crawlers that don’t adhere to our policies (up to billions of requests per day). As we continue to improve our detection mechanisms, we expect this number to increase. Earlier this month, we also began rolling out global rate limits for API traffic, with a second rollout phase planned for April 2026.

🔖 Wikimedia Attribution Framework

The Wikimedia Attribution Framework provides guidelines that data reusers can follow to ensure that sources remain clear, recognizable, and consistent in external contexts. Attribution is essential for fair acknowledgment and active awareness of Wikimedia’s community-driven content, and it’s also a key factor in the continued growth and sustainability of the free knowledge ecosystem. The framework exists to:

🔖 HTTP Message Signatures Directory

HTTP-MESSAGE-SIGNATURES allow a signer to generate a signature over an HTTP message, and a verifier to validate it. The specification assumes verifiers have prior knowledge of signers’ key material, requiring out-of-band key distribution mechanisms. This creates deployment friction and limits the ability to dynamically verify signatures from previously unknown signers.

This document defines:

A standardized key directory format based on JWKS for publishing HTTP Message Signatures keys.
A well-known URI location for discovering these key directories.
A new HTTP header field enabling in-band key directory location discovery.

🔖 Web Bot Auth - Cloudflare

Web Bot Auth is an authentication method that leverages cryptographic signatures in HTTP messages to verify that a request comes from an automated bot. Web Bot Auth is used as a verification method for verified bots and signed agents.

🔖 Framgments: April 14 / Martin Fowler

Understanding how to think about a problem domain by building abstractions (models) is my favorite part of programming. I love it because I think it’s what gives me a deeper understanding of a problem domain, and because once I find a good set of abstractions, I get a buzz from the way they make difficulties melt away, allowing me to achieve much more functionality with less lines of code. Cantrill worries that AI is so good at writing code, we risk losing that virtue, something that’s reinforced by brogrammers bragging about how they produce thirty-seven thousand lines of code a day.

🔖 Introducing a new citizen science nature app that’s geared towards the scientific community

Identifying weeds, checking out the pollen map, or discovering new plant life-forms are among the promising wealth of data available to users of PlantNet – a “Shazam!” for plants. Pierre Bonnet and computer scientist Alexis Joly introduced us to the digitally enhanced plant recognition application they developed.

🔖 Claude Code Running Claude Code in 4-Second Disposable VMs

Running Claude Code with full permissions inside a Docker container is a terrible idea. I did it anyway for about a week, then built something better.

Anthropic has an internal platform — people have been calling it Antspace since it got reverse-engineered from the Claude Code source — that runs AI coding tasks in isolated environments. It’s part of a vertical stack they’re building internally: intent goes in, code comes out, and the agent never touches the host machine.

I wanted that. Not the whole platform-as-a-service thing, just the core idea: give Claude Code a prompt, let it run with zero permission restrictions, stream the output back, grab any files it created, and destroy everything when it’s done. On a single Linux box sitting in my office.

🔖 On recognizing the handiwork of AI

As AI-generated images and texts proliferate, people have developed techniques for identifying them using clues like misshapen hands in images or distinctive words in text. This commentary situates these emerging practices within what Carlo Ginzburg called the “conjectural paradigm”: a mode of knowing that links contemporary AI detection to older traditions of medical symptomatology, art historical connoisseurship, and detective work. Yet unlike the stable or slowly evolving clues of earlier conjectural practices, the signifiers of AI involvement are rapidly shifting. This instability has consequences not only for how texts are read but also for how they are written. Authors now navigate a landscape of suspicion where their words may be misrecognized as machine generated. Rather than resolving into stable literacies, our efforts to recognize AI’s handiwork reveal the deeper uncertainties of authorship and interpretation.

🔖 We Found a Ticking Time Bomb in macOS TCP Networking - It Detonates After Exactly 49 Days

Every Mac has a hidden expiration date. After exactly 49 days, 17 hours, 2 minutes, and 47 seconds of continuous uptime, a 32-bit unsigned integer overflow in Apple’s XNU kernel freezes the internal TCP timestamp clock. Once frozen, TIME_WAIT connections never expire, ephemeral ports slowly exhaust, and eventually no new TCP connections can be established at all. ICMP (ping) keeps working. Everything else dies. The only fix most people know is a reboot. We discovered this bug on our iMessage service monitoring fleet, reproduced it live on two machines, and traced the root cause to a single comparison in the XNU kernel source.

🔖 Stanford Claude Code

Instructions for setting up claude-code to talk to the Stanford AI Playground API.

🔖 The peril of laziness lost

Left unchecked, LLMs will make systems larger, not better — appealing to perverse vanity metrics, perhaps, but at the cost of everything that matters. As such, LLMs highlight how essential our human laziness is: our finite time forces us to develop crisp abstractions in part because we don’t want to waste our (human!) time on the consequences of clunky ones. The best engineering is always borne of constraints, and the constraint of our time places limits on the cognitive load of the system that we’re willing to accept. This is what drives us to make the system simpler, despite its essential complexity

🔖 GPD Pocket

Most successful business people have one MacBook or Surface. because they not only have a stylish and gorgeous appearance but also are light and thin. Yet, their disadvantage is that they are not portable.We believe that future laptops shall not only be thin but also be small.GPD Pocket .GPD Pocket is such a product. It is not only gorgeous, ultra-light, ultra-thin like MacBook but also very small and can be taken away in a pocket at any time like a cell phone!

Digital Library Federation: DLF Webinar: Content Authenticity and Provenance in the Age of Artificial Intelligence

2026-04-16T16:46:14+00:00

On Thursday, April 15, 2026, DLF hosted Joshua Sternfeld, Independent Scholar, and Kate Murray, Digital Projects Coordinator at Library of Congress for a discussion about content authenticity, provenance, and the future of trust in libraries, archives, and museums, building on their report, Content Authenticity and Provenance in the Age of Artificial Intelligence: A Call-to-Action for the LAMs Community. The conversation explored the challenges and opportunities presented by generative AI for cultural memory institutions and considered how LAM professionals can apply the report’s call-to-action pillars to emerging real-world examples across the field.

Recording | Slides (Call to Action Presentation Slides, Claude-Generated Report: Jenny Lind Reviews)

Content Authenticity and Provenance in the Age of Artificial Intelligence: A Call-to-Action for the LAMs Community (Report)

The post DLF Webinar: Content Authenticity and Provenance in the Age of Artificial Intelligence appeared first on DLF.

Open Knowledge Foundation: Introducing OKFN’s AI Learning Labs

2026-04-15T19:32:27+00:00

We are partnering with organisations on high-impact social issues to experiment with AI and develop replicable methods together; you can join us

The post Introducing OKFN’s AI Learning Labs first appeared on Open Knowledge Blog.

In the Library, With the Lead Pipe: Another Way of Knowing: Resisting Eugenic Propaganda Through Community Archiving

2026-04-15T16:10:29+00:00

In Brief: How do information workers resist the creation of archival “deathworlds”? With rising eugenicist rhetoric in the United States, sites of cultural memory face devastating impacts. These consequences are particularly felt by Disabled and multiply-marginalized communities. This article draws on Disability Justice principles and necropolitical framings to investigate how processes of erasure can be interrupted through active collaboration and critical reevaluation of power-sharing. By supporting alternative forms of knowledge sharing and honoring the lived experience of historically marginalized communities, especially those who have faced forced institutionalization, we hope to craft alternative methodologies that center community involvement and self-determination.

By Jess Petrazzuoli-Gallagher and Ashten Vassar-Cain

Introduction

As early-career community archivists based in the United States, we are entering the archival profession at a time of fracture, lack of funding, communication breakdown, and heightened awareness of our field’s interconnectedness with policy and power. As of recently, most of the news centering Libraries, Archives, and Museums (LAMs) revolves around fear and censorship—a new list of banned books, another exhibit removal, and persistent retaliation from the Trump administration in the form of resource cuts and smear campaigns targeting institutions that refuse to bend to their will. Like many of our colleagues, we feel an overwhelming sense of urgency, guided by the weight of unanswered questions.

Reflecting on our positionality, both authors of this article are Queer and Disabled. We come to this work from research backgrounds, studying the American Eugenics movement and the use of medicalization to justify violence against marginalized bodies. Our work places us between multiple streams of knowledge. On the one hand, we are students in an ALA accredited library program, meaning we have the support and formalized training as a result of our proximity to an institution. However, most of our professional work exists outside of academia. As community archivists and activists, our work is inherently relational. It is an iterative series of mistakes, reinvention, and stories shared around tables. Our work also carries with it the lived experience of navigating ableism and violence in our daily lives, including our own experiences of institutionalization and abuse. In our practice, we reject prioritizing knowledge gained in a classroom over knowledge gained through active listening, experience, and engagement. We recognize that Disabled people have had their authority as “knowers” and knowledge producers challenged (Fricker 2007). Susan Wendell describes how “disabled people’s knowledge is dismissed as trivial, complaining, mundane (or bizarre)” (Wendell 120). Our ability to physically access knowledge is often similarly disregarded, as inaccessible buildings further restrict the ability of people with disabilities to participate as full contributors to knowledge. Because of this, we turn toward “cripistemology,” coined by Merri Lisa Johnson and Robert McCruer (2014), as an alternative to academic forms of knowledge.

Before we begin examining the process of eugenic violence and the ways it is continually recreated in our political and professional lives, we want to acknowledge that confronting this violence is far from impersonal. It is often a painful journey, especially for practitioners and community members who have been historically targeted, and those who continue to suffer harm. We struggle with the popular notion that the challenges we are currently experiencing are unprecedented. Rather, we are seeing a reinvigorated commitment to eugenicist rhetoric and policy, which have always been part of the American landscape and are enshrined in our social politics. This article examines our role as community memory workers in bearing witness and interrupting harm.

Our work is guided by Disability Justice principles, scholarship, and activism that moves beyond “rights-based” framings and toward collective action and rejection of all forms of oppression, domination, and exploitation. Because we see Disability as a dynamic axis of politics and identity, we choose to capitalize it in this article when it is used to refer to an identity category rather than as a descriptor. As Leah Lakshmi Piepzna-Samarasinha states in Care Work: Dreaming Disability Justice, “I don’t want to be fixed, if being fixed means being bleached of memory, untaught by what I have learned through this miracle of surviving. My survivorhood is not an individual problem. I want the communion of all of us who have survived, and the knowledge” (Piepzna-Samarasinha 239).

Our positionality informs our ways of documenting memory. It has also led us to imagine and create interventions that challenge power structures within our own archival practice. Through our work with the Pennhurst Memorial & Preservation Alliance’s (PMPA) Community Archive and Special Collections, we are undertaking efforts to document narratives from the self-advocacy movement led by individuals with Intellectual and Developmental Disabilities in a way that prioritizes original voice, honors lived experience, and expands access.

The Modern Eugenics Movement

In the late nineteenth century and early twentieth century, eugenics was seen as a “scientific” approach to control human genetics by limiting reproduction, resulting in forced sterilization, segregation, and systemic abuse and neglect. Eugenics and its connection to scientific racism were used to justify mistreatment on the basis of “perceived impairment.” Eugenics is one way that white supremacy violently asserts itself, claiming a scientific basis for settler colonialism and the expansion of empire. Sociologist Irving Kenneth Zola explained how medical authorities participate in enforcing state violence. In his 1972 essay “Medicine as an Institution of Social Control,” he writes that

the labels health and illness are remarkable ‘depoliticizers’ of an issue. […] By the very acceptance of a specific behaviour as an ‘illness’ and the definition of illness as an undesirable state, the issue becomes not whether to deal with a particular problem, but how and when. Thus the debate over homosexuality, drugs or abortion becomes focused on the degree of sickness attached to the phenomenon in question or the extent of the health risk involved. And the more principled, more perplexing, or even moral issue, of what freedom should an individual have over his or her own body is shunted aside. (500)

The understanding that medicine can be used as a tool to promote settler colonialist aims is sometimes referred to as Medical Imperialism (Schreier). The eugenics projects enacted by the United States resulted in over 70,000 forced sterilizations in the 20th century, though this number is likely larger, as many sterilizations have been performed without the informed consent of the individuals who had been subjected to the procedures. This did not happen overnight. It began with extensive campaigns by members of various scientific communities that captured the interest of policy makers, physicians, educators, and the American public. Eugenics bounced amongst America’s intellectual circles, ingraining itself in medical practice and scholarship. Detailed by Edwin Black in War Against the Weak, the saying “the taint is in the blood” became a prominent precept of the early eugenicists, who claimed that eradication of “undesirable traits” would result in a collectively superior “race,” and thus, collective peace and safety (Black 25). Scholar Marius Turda emphasizes a similar sentiment that capitalizes on the self-styled scientific theory of human betterment and planned breeding that eugenicists embraced. In posing biological purity as the nation’s responsibility, “eugenicists dissolved aspects of the private sphere, by scrutinizing and working to curtail reproductive, individual, gender, religious and indigenous rights. The boundary between the private and public spheres was blurred by the idea of public responsibility for the nation and the race, which came to dominate both” (2471). Such widespread influence on the “biological deterioration” of the human race captured politicians, doctors, scientists, lawmakers, and educators around the globe, and inspired horrific campaigns of genocidal violence.

The American obsession with surveillance and censorship weaponizes an idealized nuclear family, just as proponents of the early American Eugenics movement did. While libraries, archives, and museums contend with the removal of exhibits, Disability communities fear removal from public life, citing escalations that target community living protections.

In the latest iteration of American Eugenics, the Trump administration has waged a multipronged attack against Disabled Americans. In addition to dismantling of Diversity, Equity, Inclusion, and Access (DEIA) initiatives, the article “The Trump Administration’s War on Disability” from the Center on American Progress outlines an accelerated erosion of civil rights for Disabled Americans, including:

Weakening the government’s ability to enforce civil rights protections and investigate discrimination cases
Threatening access to benefits, affordable healthcare, and resources such as social services and community-based living supports
Divesting from public health infrastructure amidst an ongoing pandemic that disproportionately affects Disabled people
Decreasing employment protections related to disability
Attacking public education services offered to Disabled children
Robert F. Kennedy Jr.’s determination to “fight against Autism,” and proposed surveillance of Autistic individuals.

Actions taken by the Trump administration are most prominently seen in the July 24th executive order titled “Ending Crime and Disorder on American Streets,” which calls on the Attorney General and Health and Human Services Secretary to

enforce, and where necessary, adopt, standards that address individuals who are a danger to themselves or others and suffer from serious mental illness or substance use disorder, or who are living on the streets and cannot care for themselves, through assisted outpatient treatment or by moving them into treatment centers or other appropriate facilities via civil commitment or other available means, to the maximum extent permitted by law. (Trump, Executive Order 14321)

In combination, these measures constitute a return to “Ugly Laws,” a series of policies spanning from 1880s-1970s that removed Disabled people from public life through incarceration on the basis of “disfigurement.” Historical ugly laws and eugenics legislation reveal two intersecting dimensions of marginalization that encompass the visceral discomfort of a viewing public and the pathologization of “subnormality.” Ugly laws were then used as one metric for policing disabled, poor, and people of color for being in public, heavily relying on them being perceived as “dangerous,” “immoral,” or “unsightly”. Under Ugly Laws (1867-1974), almshouses acted as an alternative sentencing for “unsightly beggars” and “physically unable persons”—marking both people in these categories as unworthy of participating in public life, and instead subject to management by the state. State institutions for people with disabilities acted as an expansion of this carceral system, funneling individuals between its iterations, where commitment could extend to the end of that individual’s lifetime (Schweik).

Eugenic rhetoric is also employed in Executive Orders that target LAMs, invoking language of “sanity” and likening reparative descriptions to cognitive distortions (Trump, Executive Order 14253). As information workers, our choices in language, curation, and collaboration have the potential to accelerate the process of erasure and create lasting consequences for the communities depicted in the records we steward.

The Necropolitical Landscape of Memory and LAMs

Achille Mbembe, an anti-colonial Cameroonian scholar, coined the term “necropolitics.” Necropolitics expands on Foucault’s “biopolitics” through a Fanonian lens, anchoring it in opposition to apartheid and occupation. Necropolitics explores notions of oppression and mortality, while emphasizing the role of sovereignty as “the capacity to define who matters and who does not, who is disposable and who is not” (Mbembe 2003 27). Mbembe critically examines sovereignty in relation to biopower, stating that war is how nations exercise sovereignty, enact subjugation and uphold colonialism. He imagines politics as a “form of war” and asks “what place is given to life, death, and the human body (in particular the wounded or slain body)? How are they inscribed in the order of power?” (Mbembe 2003 12). Mbembe describes the creation of “deathworlds,” or “new and unique forms of social existence in which vast populations are subjected to living conditions that confer upon them the status of the living dead” (Mbembe 2003 39-40).

Some deathworlds have physical boundaries, identified by checkpoints and walls. They may take the form of state institutions that house people with disabilities, immigration detention centers, or whatever new construction of the carceral imagination best serves the goals and aims of those in power. They can also be less visible to the naked eye, consumed in the form of “small doses” of death that slowly erode and constrict our personhood. Necropolitics is wielded as “the power to manufacture an entire crowd of people who specifically live at the edge of life, or even on its outer edge — people for whom living means continually standing up to death” (Mbembe 2003 37-38).

Museums and archives occupy critical positions in this necropolitical landscape, serving as repositories of historical evidence and active creators of cultural memory. Our institutions have the power to enact “archival death” through our curatorial choices. Archival life exists in the tension between preservation and interpretation, between fixed materiality and fluid meanings attributed across time. Archival death becomes particularly insidious through its apparent neutrality—the removal of exhibits, deaccessioning of materials, and reframing of narratives—while achieving the same erasure as more overt forms of violence.

Prominent discussions in the field of digital stewardship and archival processing ask us if we are ready to confront the fact that our professional practices have upheld and facilitated the under-documentation and erasure of people from Black, Indigenous, Immigrant, Disabled, and Queer communities from the historical record and what inclusive forms of archiving might look like within our field (Duff 124).

“Archival Death” and Curatorial Power

Archival literature and practice relating to disability has historically focused on medical narratives, accessible practices, and histories authored by those in power rather than the ways in which disability and marginalized populations are documented. These framings often result in a narrow, medicalized representation of people within our collections that fails to capture the complex, intersectional lives of disabled people. Primarily residing in medical libraries and government archives, practices related to archiving disability tend to focus heavily on the interpretation of disability in the context of science and medicine, relying on labels prescribed by medical authorities rather than the social and embodied experience of disability. Kelvin White’s overview of the genesis of the field of archival preservation in “Promoting Reflexivity and Inclusivity in Archival Education, Research, and Practice,” provides valuable insight into the field’s standardized practices that have been shaped by people in positions of power—mainly those who were not from historically marginalized communities themselves. (K. White 117).

The intersectional lives of people with disabilities and their interactions with medical violence and eugenics are erased, and the physical archives remain largely inaccessible to disabled patrons and disabled professionals. The record may technically survive, but the life within it does not. Archival death, then, is not only about what is discarded or deaccessioned. Instead, it operates equally through how/what remains and if it is actively hidden.

According to Tobin Siebers, disability can be seen as an “elastic social category” that changes depending on the social context, making a singular definition for archival usage difficult. Without a proper theoretical framework like complex embodiment —which Sara White explains as evaluating disability as an experience—archivists and political actors risk inserting their internal biases about disability, eugenics, and state-sanctioned violence against marginalized communities (S. White).

Beth Linker’s “On the Borderland of Medical and Disability History: A Survey of the Fields,” expands on White’s historical discourse and links the rise of the American medical system to the documentation of disease-centric medical history that often neglects disability. The academic study of medical history in the U.S. began in the early 1930s—much like the beginning of the professional archival field—largely due to émigré scholars who had been trained in medicine and the humanities in German-speaking Europe. These scholars, including Henry Sigerist, Owsei Temkin, and Erwin Ackerknecht (all of whom spent time at Johns Hopkins), brought with them a deep commitment to continentally-infused theories and ideas, particularly concerning disease. As a result of this intellectual background and the research interests of these influential figures, disease history became the central research aim of the newfound field of medical history in the United States. This focus was not predetermined but rather a product of the time, as the individuals who shaped the discipline were medical practitioners. The relative neglect of disability within medical history contributed to the emergence of disability history as a distinct field in the late 1990s. “New” disability historians explicitly defined their discipline in contrast to medical history, arguing that the divergences between the two fields defined disability history’s parameters. Disability historians argued that the medical model defined disability solely as a consequence of biological factors—such as congenital or chronic illness, injury, or deviation from a perceived “normal” biomedical structure or function—and seeks to “fix” or cure its effects at the individual level.

When disability materials are collected solely through a medical lens—cataloged under disease categories, described in clinical language, stripped of the political and social contexts that gave them meaning—the person is rendered a faceless patient rather than an agent. When it comes to records and legacies of state-run institutions, such as the Fernald State School and Hospital in Massachusetts or the Pennhurst State School and Hospital in Pennsylvania, sketches of the institutions themselves are widely available. The institutions are venerated, retaining a life of their own even after the buildings have been shuttered. However, former residents are reduced to gaps in the historic record. Family members and researchers requesting information about their lives are met with barriers such as incomplete or lost records, restricted access to records, or lack of resident perspectives. In a way, the archive may act akin to the institution itself, dressing harm and isolation in the language of “care,” and participating in epistemic injustice against people with disabilities and their many “ways of knowing.”

In “Documenting Disability History in Western Pennsylvania,” Bridget Malley highlights Helen Samuels’ documentation strategy as “well designed to address gaps in the historic narrative by ‘provid[ing] a useful framework for discussing selection issues,’” particularly with respect to marginalized communities (17). Documentation strategy as methodology “guides selection and assures retention of adequate information about a specific geographic area, a topic, a process, or an event that has been dispersed throughout society” (17). Documentation strategy emphasizes structuring the inquiry and examining the form and substance of the available documentation. This involves actively seeking to understand the documentary universe related to a topic, identifying where documentation exists and, crucially, where it doesn’t exist.

Traditional appraisal practices can inadvertently contribute to archival gaps by reflecting the biases and perspectives of the appraisers, leading to the erasure of marginalized voices like those of people with disabilities. Documentation strategy aims to move away from subjectivity by incorporating a deeper understanding of the topic and the perspectives of those involved. For Malley, the collaboration between Western Pennsylvania Disability History Action Consortium (WPDHAC) members (community experts) and archivists from the Heinz History Center allowed for a merging of archival and community knowledge, leading to more nuanced appraisal decisions and potentially filling gaps based on community-identified needs in the archives.

Archival death was built into the profession’s foundations: the same apparatus that preserved evidence of national progress systematically failed to collect, or actively destroyed, evidence of the state’s violence against its most marginalized citizens. As uncomfortable as it may be to confront, this was not an oversight. It was a feature of archives designed to construct a cohesive national narrative by using seemingly neutral representations of history that avoided critical insights into the past, and something that we must take active steps to remedy. The records relating to institutionalized individuals that do survive from this era—asylum logs, sterilization orders, commitment papers—were authored by perpetrators, not survivors. Even when disabled people appear in the archive, they appear as objects of intervention rather than witnesses to their own lives. When disability materials are appraised, arranged, and described according to the medical model, the archive reproduces the very framework that justified confinement and cure. The disabled person is preserved as a case, not a life; a diagnosis, not a history. This is erasure through categorization—effective in determining who the ‘dead’ and the ‘living’ might become to future researchers, advocates, and communities seeking to understand what was done and what survived.

Community-Led Collecting: Alternative Methods for Protecting Shared History

Community-based archives have emerged as crucial alternatives to mainstream institutions, particularly for documenting the histories of marginalized groups. These archives are often created by and for the communities they serve, prioritizing agency, self-determination, and the preservation of narratives that challenge dominant frameworks. Community archives intentionally subvert the “neutrality” of institutional preservation by centering the values, priorities, and privacy concerns of their communities. Community archives may employ participatory appraisal and curation strategies, working directly with community members to identify, select, and describe materials. This approach seeks to capture the richness and diversity of lived experience, rather than reducing disability or other markers of identity to a social problem. It also reflects the interconnectedness of “ways of knowing.”

In a 2022 interview, Achille Mbembe explains “the epoch we have entered into is one of indivisibility, of entanglement, of concatenations. Times of concatenation presuppose that our bodies have become repositories of different kinds of risks” (Mbembe 2022). Risk is very present in discussions of disability politics. “Risky bodies,” as described by Hi‘ilei Julia Kawehipuaakahaopulani Hobart and Tamara Kneese, are subjected to coercive forms of care (Hobart & Kneese), under the assumption that those living in the intersections of positionality cannot be trusted as knowers. One path away from paternalistic and colonial imaginings of care and knowledge is community-led intervention.

Some notable sites of community intervention we have encountered in our work include the Living Archives on Eugenics (LAE), The Anti-Eugenics Collective at Yale, and the From Small Beginnings Collective. The Living Archives on Eugenics in Western Canada (LAE) provides us a realistic vision for the future of community archival practices centered on disability and survivors of the eugenics movement. Working directly with survivors the project “raised awareness of historical and contemporary manifestation of eugenics [by capturing and disseminating] survivor’s stories.” Interactive collections provide historical context to the eugenics movement in Western Canada and emphasize the bond that was created between curators, archivists, and survivors. The Anti-Eugenics Collective at Yale situated Yale’s campus as the former headquarters of the American Eugenics Society, and their related collections as a site of harm and opportunity for reparation. They engage this troubled history through workshops with K-12 educators, students, medical professionals and the general public. The global collective From Small Beginnings is a group of anti-eugenics activists that helps educators and researchers learn of ongoing efforts to disrupt eugenics in action, and combats isolation by building a network of committed individuals and organizations. These projects meaningfully balance confrontation and collaboration through outreach and information organization.

These initiatives, coupled with the work done by the Disability Archives Lab on centering critical disability studies in archival research and practice, and the recent publication Preserving Disability: Disability and the Archival Profession (Brilmyer & Tang, 2024), inspired us to think about potential places of intervention in our own archival process. We established the Pennhurst Memorial & Preservation Alliance Community Archives in 2024 with support from the Pennsylvania Historic & Archival Records Care (HARC) grant, organizing materials collected by the organization over a span of approximately five decades. Despite being a volunteer-run organization, we wanted to ensure that we could make our records, largely generated by self-advocates with Intellectual and Developmental Disabilities, accessible to the public. Of particular significance is the archive’s Speaking for Ourselves (SFO) collection, which documents the critical role of self-advocacy organization Speaking for Ourselves in the disability rights movement emerging from state institutions in Pennsylvania. By centering self-advocates’ political struggle, agency, expertise, and vision for the future, the archive challenges dominant narratives of Disability, medicalization, and victimhood.

Our process required that we confront power imbalances and collective trauma surrounding documentation and consent. In our previous archival research and practice, we noticed that academic and state archives privileged the voices of medical and institutional authorities. Disabled individuals, especially people with Intellectual and Developmental Disabilities, were excluded from authoring knowledge, instead cast as subjects of research. Our collections contained a unique perspective that was absent from more “traditional” state and academic archives, acknowledging People with Intellectual and Developmental Disabilities as originators of our records and contributors to social change.

After over a year of sitting in on community board meetings and listening to requests from the community for long-term preservation and digital accessibility, we applied for funding through the Council on Library and Information Resources’ “Digitizing Hidden Collections: Amplifying Unheard Voices” grant program, and were awarded funds to run a two year digitization project. As grantees of this program, we seek to digitize and make accessible over 9,000 items within our collections.

In order to keep us accountable to our mission and the community’s requests, we have implemented a series of strategies to aid us in our intervention. To us, this means prioritizing audio and visual material for digitization. We recognize that many of the self-advocates who originated the materials in our collections did not participate in traditional forms of communication or written record keeping, instead relying on dictation, audio and video recording, and other accessible modes of knowledge sharing and creation In an attempt to preserve original voice, we are engaging in consultation with surviving advocates depicted in our materials. We are also planning for quarterly access consultations with users across disability communities. These consultations will allow us to continually integrate feedback and ensure that the open access metadata and digital exhibits generated through this project are accessible to the widest possible user base.

Planning this project required us to reimagine what our role as archivists might look like. We understand that the self-advocates who originated and are depicted in the collections have had much of their lives documented without consent in the form of institutional medical records. To avoid replicating similar harm, we have decided to forgo a traditional “donor” model, in favor of a “stewardship agreement.” The terms of this agreement allow our archive to take actions necessary for preservation, make collections publicly available, and provide open access to associated metadata, with the understanding that the physical materials are property of Speaking for Ourselves.

We are documenting each step of our process so that we can contribute findings from this project to the larger Disability archival community, so that it may be replicated and expanded upon. Additionally, we designed the project to prioritize intergenerational collaboration, bringing younger members of our community in conversation with the historical context for our current struggles and learning from the voices of Disabled elders. Over the next two years, we are engaging in an iterative learning process, in hopes of fulfilling the request for digital access and making the perspectives and activism of self-advocates known to a wider audience.

Interrupting Erasure: How Archivists Protect the Future

Critically rethinking power and ownership requires institutions and practitioners to develop new methodologies that recognize diverse community members as primary stakeholders, and archives as sites of evolution and growth. This involves moving beyond inclusion toward genuine power-sharing and decolonial praxis. The archive becomes a site of active creation and a path to alternative “ways of knowing,” which in turn resist the creation of “deathworlds.” In “Out of the Dark Night: Essays on Decolonization,” Achille Mbembe affirms that

humanity is to be made to rise [faire surgir] through the process by which the colonized subject awakens to self-consciousness, subjectively appropriates his or her I, takes down the barrier, and authorizes him- or herself to speak in the first person. This awakening and appropriation aim not only at the realization of the self, but also, more significantly, at an ascent into humanity, a new beginning of creation, the disenclosure of the world. (62)

Many Disabled self-advocates have communicated distrust toward record-keeping solely dictated by medical and legal authorities. Disability community archives can demonstrate one avenue for grassroots preservation efforts that maintain disabled people’s ownership/autonomy over their narratives of survival and liberation, challenge dominant medical narratives that have historically justified confinement, and resist archival death.

The formation of the archival profession and of many LAM institutions are tied to colonial wealth and power. Even after constant re-evaluation of practices and strategic upheaval, LAMs come in contact with a different type of “deathworld” on a daily basis, and participate in decisions that determine what – and who– is remembered. As librarians, archivists, and museum professionals, we are responsible in part for making difficult choices with the collections we steward. We are also responsible to the communities depicted in those records. Speaking of Zola’s writings on medicalization, he cautions that “not only is the process masked as a technical, scientific, objective one, but one done for our own good” (502). Though not medical professionals, information workers risk modeling a similar attitude toward disability in our work, and participating in archival erasure under the guise of objective processes. Even as the risk landscape changes, we must recognize that our roles are not neutral. To combat erasure, we must take an active role in interrupting the weaponization of memory against the most marginalized. Interruption can take many forms, and requires ongoing reflection and adaptation.

Drawing on the practical work of building and sustaining the PMPA Community Archive, we are energized in imagining how LAM professionals can engage in active forms of resistance. The sustainability of grassroots archival work depends on the active partnerships between individuals, communities, larger institutions, and solidarity networks across libraries, archives, and museums. Without webs of mutual aid and professional collaboration, we all remain vulnerable to the same political pressures that have compromised larger institutions.

Dealing with histories of medical violence mandates space to grieve. Leah Lakshmi Piepzna-Samarasinha invites grief as an active part of the process, arguing

that feelings of grief and trauma are not a distraction from the struggle. For example, transformative justice work—strategies that create justice, healing, and safety for survivors of abuse without predominantly relying on the state—is hard as hell! What would it be like if we built healing justice practices into it from the beginning? (42)

The stakes of this work extend beyond professional practice. It challenges all of us to consider whose lives matter in the American memory. The PMPA Community Archives imagines community-controlled historical preservation as a form of active survival and maintains originators’ ownership over their narratives of resistance and liberation—narratives that are urgently felt when policies seek to industrialize age-old structures of abuse and reintroduce the ‘legal removal’ of disabled people from the public eye.

Acknowledgements

We are deeply grateful to our Lead Pipe editors, Jess Schomberg and Pam Lach, whose feedback and support helped shape this article. We are especially grateful to our reviewer, Gracen Brilmyer, whose work we greatly appreciate and respect. This work emerged from and belongs to the community of self-advocates and activists who have shaped the Pennhurst Memorial & Preservation Alliance. Their decades of organizing, documenting, and demanding recognition created the conditions for this scholarship to exist. Their lived expertise, activism, and commitment to truth-telling made this research possible. We are grateful to be entrusted with carrying forward their labor of memory and resistance. Any insights here reflect their collective work.

Works Cited

Assistant Secretary for Public Affairs (ASPA). “Secretary Kennedy Appoints New Interagency Autism Coordinating Committee to Advance Fight Against Autism.” HHS.Gov, 28 Jan. 2026, www.hhs.gov/press-room/hhs-kennedy-appoints-new-interagency-autism-coordinating-committee.html.

Black, Edwin “America’s National Biology,” in War Against the Weak: Eugenics and America’s Campaign to Create a Master Race Dialog Press, 2012, pp. 21-42.

Brilmyer, Gracen, and Lydia Tang, editors. Preserving Disability: Disability and the Archival Profession. Library Juice Press, 2024.Disability Archives Lab, disabilityarchiveslab.com.

Duff, Wendy, et al. “Investigating the Impact of the Living Archives on Eugenics in Western Canada.” Archivaria: The Journal of the Association of Canadian Archivists, 2019, archivaria.ca/index.php/archivaria/article/download/13701/15099.

Eugenics and its Afterlives, www.antieugenicscollective.org.

Fricker, Miranda. Epistemic Injustice: Power and The Ethics of Knowing, Oxford University Press, 2007.

FromSmallBeginnings, www.fromsmallbeginnings.org.

“Hidden Collections.” CLIR, 28 Oct. 2025, www.clir.org/hiddencollections.

Hobart, Hi‘ilei Julia Kawehipuaakahaopulani, and Tamara Kneese, editors. Radical Care: Survival Strategies for Uncertain Times. Duke University Press, 2020.

Ives-Rublee, Mia, and Casey Doherty. The Trump Administration’s War on Disability. Center for American Progress, www.americanprogress.org/article/the-trump-administrations-war-on-disability.

Johnson, Merri Lisa and Robert Mcruer. “Cripistemologies: Introduction.” Journal of Literary & Cultural Disability Studies 8 (2014): 127 – 147

Linker, Beth. “On the Borderland of Medical and Disability History: A Survey of the Fields.” Bulletin of the History of Medicine, vol. 87, no. 4, Dec. 2013, pp. 499–535, https://doi.org/10.1353/bhm.2013.0074.

Living Archive on Eugenics, www.eugenicsarchive.ca.

Malley, Bridget. “Documenting Disability History in Western Pennsylvania.” The American Archivist, vol. 84, no. 1, 1 Mar. 2021, pp. 13–31, https://doi.org/10.17723/0360-9081-84.1.13.

Mbembe, Achille. “Necropolitics.” Public Culture, vol. 15, no. 1, 1 Jan. 2003, pp. 11–40, https://doi.org/10.1215/08992363-15-1-11.

Mbembe, Achille. Necropolitics. Duke University Press, 2019.

Mbembe, Achille. Out of the Dark Night: Essays on Decolonization. Columbia University Press, 2021.

Mbembe, Achille. “Achille Mbembe: Planetary Politics for All Creation.” Interview by Noema Magazine. Noema Magazine, 11 Jan. 2022, Noema Magazine.

NPR. “The Supreme Court Ruling That Led to 70,000 Forced Sterilizations.” Fresh Air, 7 Mar.
2016, www.npr.org/sections/health-shots/2016/03/07/469478098/the-supreme-court-ruling-that-led-to-70-000-forced-sterilizations.

Pennhurst Memorial & Preservation Alliance, preservepennhurst.org.

Pennsylvania State Archives. Historical & Archival Records Care Grant Program, Commonwealth of Pennsylvania, www.pa.gov/services/phmc/apply-for-the-historical—archival-records-care-grant-program.

Piepzna-Samarasinha, Leah Lakshmi. Care Work: Dreaming Disability Justice. Arsenal Pulp Press, 2021.

Schreier, H. & Berger, L. (letter), ‘On medical imperialism’, Lancet, 1:1161, 1974.

Schweik, Susan M. The Ugly Laws: Disability in Public. New York University Press, 2010.

Sins Invalid. “10 Principles of Disability Justice.” sinsinvalid.org/10-principles-of-disability-justice.

Speaking For Ourselves, speaking.org/.

Turda, Marius “Legacies of Eugenics: Confronting the Past, Forging a Future,” Ethnic and Racial Studies 45, no. 13, 2022, pp. 2470–77, https://doi.org/https://doi.org/10.1080/01419870.2022.2095222.

United States, Executive Office of the President [Donald Trump]. Executive Order 14253: Restoring Truth and Sanity to American History, 28 March, 2025, www.whitehouse.gov/presidential-actions/2025/03/restoring-truth-and-sanity-to-american-history/.

United States, Executive Office of the President [Donald Trump]. Executive Order 14321: Ending Crime and Disorder on America’s Streets, 24 July 2025, www.whitehouse.gov/presidential-actions/2025/07/ending-crime-and-disorder-on-americas-streets/.

Weinberg, Hannah. “Tracking the Trump Administration’s Attacks on Libraries.” American Libraries Magazine, 1 May 2025, americanlibrariesmagazine.org/2025/03/19/tracking-the-trump-administrations-attacks-on-libraries.

Wendell, Susan “Toward a feminist theory of disability.” Hypatia, 1989.

White, Kelvin L., and Anne J. Gilliland. “Promoting Reflexivity and Inclusivity in Archival Education, Research, and Practice.” The Library Quarterly, vol. 80, no. 3, July 2010, pp. 231–248, https://doi.org/10.1086/652874.

White, Sara. “Crippling the Archives: Negotiating Notions of Disability in Appraisal and Arrangement and Description.” The American Archivist, vol. 75, no. 1, Apr. 2012, pp. 109–124, https://doi.org/10.17723/aarc.75.1.c53h4712017n4728. Zola, Irving Kenneth. “Medicine as an Institution of Social Control” The Sociological Review, 20: 487-504. https://doi.org/10.1111/j.1467-954X.1972.tb00220.x

Web Science and Digital Libraries (WS-DL) Group at Old Dominion University: 2026-04-15: ACM Capital Region Celebration of Women in Computing (CAPWIC 2026) Trip Report

2026-04-15T13:48:16+00:00

This year the ACM Capital Region Celebration of Women in Computing (CAPWIC 2026) was held from March 27–28 as an in-person event. The conference took place in Alexandria, Virginia, and the event was hosted by the Virginia Tech's Institute for Advanced Computing (IAC). CAPWIC is all about bringing together women in computing and their peers to support each other and grow in the field. The conference connects students, faculty, and industry professionals from across the Capital Region – Pennsylvania to Virginia to share ideas, discuss research, and build a strong, supportive community.

The conference featured workshops, technical talks, flash talks, research shorts, poster presentations, as well as panel, birds-of-a-feather and keynote sessions. I was the only participant from Old Dominion University's Web Science and Digital Libraries (WS-DL) research group this year, where I presented a research short. The event included parallel sessions across various categories and topics, and I attended sessions from each category.

Conference Venue: Institute for Advanced Computing (IAC), Virginia Tech, Alexandria, Virginia

Day 1: March 27, 2026

The first day of the conference began with a campus tour and a graduate/career fair. Day 2 also included the campus tour and graduate/career fair for those who missed. This was followed by opening remarks and dinner. Next, the first keynote was delivered, and the day concluded with closing remarks for Day 1.

Campus Tour: Drone Display and Immersive Visualization Lab Visit

The Institute for Advanced Computing (IAC) of Virginia Tech is a research institute located at Alexandria, Virginia. The institute offers hands-on learning opportunities for graduate students in computer science and computer engineering. Specialized labs are available at the institute for research in immersive visualization, drone technology, wireless, quantum, and brain-inspired computing systems. We got the opportunity to visit the Drone Lab and Immersive Visualization Lab.

The Drone Lab featured an indoor drone cage used to conduct flight experiments in a controlled and safe environment. The lab team introduced us to the fundamentals of unmanned aerial system technology and shared insights into their ongoing research. One of the interesting discussions was about how they are trying to detect commercial off-the-shelf (COTS) drones, which can be used for attacks or unauthorized surveillance. They also gave us a chance to fly drones inside the cage, which was a really fun and thrilling experience.

The Immersive Visualization Lab provided immersive projection on three walls and the floor, allowing users to be fully immersed in visual representations of data and other phenomena. We had the opportunity to experience a virtual walk through a beautiful garden, which felt truly magical. It was amazing to see how visual design and 3D modeling can bring environments to life and let us explore places we would not normally be able to experience in person.

📍Day 2 @capwic #CAPWIC2026!

💡Cool drone display and immersive lab visits at the event! 🤩#WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/VoanQtG8lh
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Graduate/Career Fair

The sponsors of the event organized a graduate/career fair for the attendees. There were representatives from ACM Women in Computing (ACM-W), Virginia Tech’s Computer Science Department, Virginia Tech’s Sanghani Center for Artificial Intelligence & Data Analytics, University of Mary Washington's College of Business, Northeastern University’s Khoury College of Computer Sciences, and Women in CyberSecurity (WiCys). They shared information about graduate programs and career opportunities in research and academia, and also provided valuable feedback on resumes. I had the opportunity to interact with several representatives, which helped me better understand potential career paths in academia and research.

📍Day 2 @capwic #CAPWIC2026!

💼 Graduate/Career Fair is going on! #WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/0UumYI7QfA
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Opening Remarks

After the campus tour and graduate/career fair, the conference began with the opening remarks from the organizers. The head of Virginia Tech’s computer science department, Christine Julien, welcomed everyone to the conference. The organizing chairs, Sehrish Basir Nizamani from Virginia Tech and ODU alumna (PhD, 2004) Mona Rizvi from James Madison University, provided an overview of the tracks for each category and schedule of the conference. The program included 2 panels, 5 workshops, 8 technical talks, 12 flash talks, 22 research shorts, 43 posters, and 1 birds-of-a-feather session, all conducted across parallel sessions.

📍Day 1 @capwic #CAPWIC2026!

📣 Opening remarks happening now by the organizers -- Sehrish Basir Nizamani and Christine Julien from @virginia_tech!#WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/9McdunVnHX
— Tarannum Zaki (@tarannum_zaki) March 28, 2026

Keynote #1: Tools in Your Toolbox: What I've Learned as a Professional Female Computer Scientist

Christine Julien introduced the first keynote speaker Laurian Vega, a Senior System Engineer at Booz Allen Hamilton. She shared her skills and expertise that she developed throughout her career as a female computer scientist. One of the key takeaways from the keynote was that no effort is ever wasted as long as you learn something from it. The speaker talked about how important it is to invest in soft skills and to build strong networks. She also encouraged us to choose workplaces that align with our values and treat us well. A point I found especially meaningful was that mental health is just as important as physical health. The speaker also emphasized caring about the work we do and using our skills to give back to the community. Finally, she highlighted that a PhD degree is not about becoming an expert in everything, but about continuously learning and growing. Overall, the talk highlighted that success in computing is not just about technical ability, but about developing a balanced toolkit that supports both personal and professional growth.

📍Day 1 @capwic #CAPWIC2026!

*Invest in soft-skills
*No effort is wasted when you learn from it
*Learn about learning
*Care about your work

💬 Laurian Vega from @BoozAllen started the first keynote with useful advice for PhD life.#WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/venFmGWkXv
— Tarannum Zaki (@tarannum_zaki) March 28, 2026

Day 2: March 28, 2026

Day 2 started off with breakfast, followed by the second keynote. Before the lunch break, two parallel sessions were held. During the first session, I attended a workshop and a technical talk. In the second session, I attended flash talks and another workshop. After the lunch break, there were two more parallel sessions. I attended a panel and a poster session for the third one and research shorts from the final one.

Keynote #2: Goodbye Imposter, Hello Winner: Overcoming Perceptual Expectations to Reclaim Excellence

Erika Olimpiew from Virginia Tech introduced the second keynote speaker Candace Aku, a Senior Technical Program Manager at Google Public Sector. She shared her journey in the tech industry and the process of creating a professional identity beyond others’ expectations. The speaker emphasized on the importance of not limiting one’s potential before even starting a career, and of reflecting on whether actions are driven by personal goals or others’ expectations. The speaker also highlighted how constantly chasing the “next” can lead to burnout, reminding the need to prioritize well-being. The discussion on the weight of expectations such as intelligence, imperfection, fear of failure, and judgment was particularly insightful, as these factors often contribute to imposter syndrome. The keynote was highly motivating, encouraging individuals to embrace their identity, challenge limiting beliefs, and grow without sacrificing their well-being.

📍Day 2 @capwic #CAPWIC2026!

💬 The second keynote was by Candace Aku from @GooglePubSector, discussing how to thrive in the computing industry while dealing with the weight of expectation and burn-outs.#WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/2WKCUTRf4y
— Tarannum Zaki (@tarannum_zaki) March 28, 2026

Workshop: Debugging Your Resume

Aubrey Baker, an eCommerce Web Developer from Red Van Workshop, and Holly Wilsey, a Video Game Engineer from Purple Basil Games, organized a hand-on workshop on preparing resumes. The session was chaired by Nguyen Ho from Loyola University Maryland. They shared best practices on customizing resumes for various purposes, such as academic work, internships, employment, and volunteer experiences. Participating in this workshop gave me a better understanding of how Applicant Tracking Systems (ATS) screen resumes before they are seen by a human. I learned how small details in formatting and wording can impact visibility, and how to avoid common mistakes that can weaken a resume. The breakout sessions were especially helpful, as we got to review and improve our resumes in a group while receiving useful feedback.

📍Day 2 @capwic #CAPWIC2026!

⚙️ Aubrey Baker from @RedVanWorkshop and Holly Wilsey from Purple Basil Games discussed about the basics of preparing resume in the workshop session.#WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/QiMxF6ABtY
— Tarannum Zaki (@tarannum_zaki) March 28, 2026

Technical Talk: Education & Inclusion

Denise D'Angelo, a Transformation Technology Leader at DynamicD Enterprises, presented a technical talk “Designing the AI-Ready Workforce” as part of the Education & Inclusion track. The session was chaired by Mohammed Farghally from Virginia Tech. The speaker offered valuable insights into how to approach AI-enabled work more thoughtfully, especially in the context of hiring. As AI becomes part of our everyday work, traditional ideas about roles and performance are evolving, influencing both opportunities and trust. She explained that being “AI-ready” is not just about knowing how to use AI tools, but about understanding how people and AI systems work together. The talk was very helpful to clearly understand how to prepare for interviews in an AI-driven hiring landscape.

📍Day 2 @capwic #CAPWIC2026!

🔌 AI is the member of the hiring team now. Denise D'Angelo from dynamicD shared useful strategies to navigate the AI hiring landscape in the Tech Talk session.#WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/wbZJ0YG1rh
— Tarannum Zaki (@tarannum_zaki) March 28, 2026

Flash Talk: Trust, Fairness, and Societal Impact of AI

Mona Rizvi chaired the flash talk session on the Trust, Fairness, and Societal Impact of AI track.

As the first presenter, Sadia Afrin Mim from George Mason University presented “LLM-Guided Input Generation for Causal Fairness Testing.” Current fairness testing methods in machine learning systems often create unrealistic test cases by ignoring how features relate to each other in real-world situations. To address this limitation, the presenter introduced a new approach that uses large language models (LLMs) to generate more meaningful and context-aware test inputs.

📍Day 2 @capwic #CAPWIC2026!

⚡The Flash Talk session started with Sadia Afrin Mim from @GeorgeMasonU, presenting how LLM's can be leveraged for fairness testing.#WomenInComputing #VT_IAC @virginia_tech #FairnessTesting pic.twitter.com/JHUu9aW1De
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Next, Arshnoor Bhutani and Mahi Sanghavi from University of Maryland, College Park presented “Data-Driven Exploration of Physiological Factors Perpetuating Bias in Pulse Oximetry Readings for At-Home Use.” They examined bias in pulse oximeters which is used to measure blood oxygen levels and found that skin tone remains an important contributor to this bias. They analyzed the BOLD data set and identified a clear pattern showing that errors increase as skin tone gets darker. Their work aims to better understand these patterns so that corrections can be developed to help reduce health disparities.

⚡Arshnoor Bhutani & Mahi Sanghavi from @UofMaryland presented how data-driven methods can be used to examine the varied bias across skin tones while controlling for physiological factors.#HealthEquityResearch pic.twitter.com/DdgYkp19u8
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Next up, Khoulood Alharthi from Virginia Tech presented “Gender, Culture, and Privacy: Navigating Social Media Concerns in Saudi Arabia.” She explored how privacy concerns on social media are shaped by culture and gender, focusing on users in Saudi Arabia. She emphasized that privacy is not only determined by platform features, but deeply influenced by users’ social and cultural values such as modesty, reputation, and social boundaries. Her work provided valuable insights into how social media platforms and privacy settings can be designed to better align with users’ cultural expectations.

⚡Khoulood Alharthi from @virginia_tech presented the importance of considering culture and gender for evaluating privacy concerns of social media. #Privacy #SocialImpact pic.twitter.com/OdcpyZlY8d
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Christopher Parham from Virginia State University presented the next flash talk, “A Trust-Aware, Biometrically-Secured Social Network Using Decentralized Identity Protocols and the Analytic Hierarchical Process for Collaboration.” His talk focused on improving security and trust in online systems by addressing human factors in cybersecurity. He proposed a novel decentralized method that creates one-time biometric features to prevent attacks like replay or misuse of credentials. His work aims to create a more secure, user-friendly authentication framework that supports reliable and trust-based collaboration.

⚡Christopher Parham from @VSU_1882 introduced a framework based on human-centered cybersecurity for a secure collaborative research environment.#Networking #DecentralizedSystem pic.twitter.com/lIAP04znUM
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Saanvi Shashikiran from Georgetown University presented the last flash talk, “Understanding State-Level AI Readiness Policy.” She explored how prepared different U.S. states are for adopting AI, focusing on the role of policies, infrastructure, and support systems. She examined five states and analyzed government documents to understand how policies regulate AI use. She discussed how text-based search methods can be used to identify policies relevant to AI readiness, and found that a specific scoring method (BM25) performed most effectively.

⚡Saanvi Shashikiran from @Georgetown presented text-based similarity approaches to track changes in policy documents that support AI readiness.#AI #StatePolicy pic.twitter.com/x9XBsbfTDH
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Workshop: Cyber Hygiene That Sticks: Research in the K-12 Space on Cybersecurity

Deborah Kariuki, an Assistant Teaching Professor at University of Maryland Baltimore County, led a workshop on how cybersecurity in K-12 education can be improved through more interactive and human-centered approaches. She emphasized that relying only on rules and one-time sessions are not enough to strengthen cyber hygiene. She demonstrated how interactive activities through real-world scenarios can help students effectively recognize threats like phishing, password safety, and data-sharing online. She also talked about the broader efforts of organizations like WiCyS in actively promoting cybersecurity education and awareness, helping students build confidence and lasting digital safety habits.

📍Day 2 @capwic #CAPWIC2026!

⚙️ Rules alone cannot support digital safety habits. Human-centered design can help to create durable cyber hygiene practices for K-12 education -- workshop by Deborah Kariuki from @UMBC.#WomenInComputing #VT_IAC @virginia_tech @WiCySorg pic.twitter.com/qHkOxfKc0T
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Panel: Navigating the Path to Grad School: Discuss, Reflect, and Make an Informed Decision

Mohammed Seyam from Virginia Tech moderated a panel session that provided a reflective perspective on what it is truly like to pursue an advanced degree and how to decide whether it is the right path. The panel featured Madison Barton, a graduate admission counselor at Northeastern University, Mohammed Farghally, a collegiate assistant professor at Virginia Tech, Promise Owa, a graduate student at Northeastern University, and Chandani Shrestha, an Assistant Professor at James Madison University. The panelists shared their own journeys, including their uncertainties, key turning points, and lessons learned, while addressing common questions such as –

Does taking a year off or doing a job before starting grad school have any impact?
What are the key factors to choose grad school? Funding? Research facilities?
How much research interest matters to sustain throughout grad school?
What are some challenges for women in CS grad school?
How to deal with imposter syndrome during grad life?

The session was very encouraging for the participants to think intentionally about their goals, interests, and readiness for grad school.

📍Day 2 @capwic #CAPWIC2026!

🗣️ Panel session on navigating the path to graduate school is happening now!

*Does taking a year off before starting grad school have any impact?

*What are the key factors to choose grad school?#WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/anIa6CSMPD
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Posters: Cybersecurity, Privacy & Responsible AI

Jessica Zeitz from University of Mary Washington chaired the poster session on the Cybersecurity, Privacy & Responsible AI track.

Fairuz Nawer Meem from George Mason University presented two posters. One poster titled “Hope or Hype? Understanding Vibe Coding through Software Practitioner Discussions” showed analysis on online discussions to understand how developers’ opinions about “vibe coding” changed over time. Another poster, “Well-Being in AI-Assisted Software Development,” showed an experimental study on how using AI tools affects developers’ stress, emotions, and overall well-being while coding, compared to coding without AI. Sadia Afrin Mim, also from George Mason University, presented the poster “Towards Practical Discrimination Testing for Software Systems,” discussing the evaluation of a user study that fairness-specific tools, along with AI support, help developers find and understand bias in software more easily.

Min Zhang from Virginia Tech proposed a way for smaller local AI models to get help from powerful remote models while protecting sensitive data in her poster “PrivacyR1: Privacy-Preserving Collaborative Reasoning in Multi-Agent Systems.” Jennifer Alexandra Thompson, also from Virginia Tech, presented the poster “Exploring Socioeconomic Status Narratives of Computer Science Students,” exploring how a student’s socioeconomic background affects their access to technology and success in computer science education. Another student from Virginia Tech, Kimberly Giordano presented the poster “Beyond the Android Manifest: Analyzing Native Libraries and Eye-Tracking Use in Virtual Reality Applications,” showing an analysis on how VR apps use eye-tracking data based on Android Manifest evaluation and native code inspection. She found that some apps may access data without clearly informing users.

Rebecca George from the College of William & Mary presented performance evaluation of a new storage system (DAOS) on a large supercomputer and identified the best ways to optimize data reading and writing for faster performance in her poster “Benchmarking DAOS Filesystem on Aurora.” Zahra Rizvi, also from the College of William & Mary, presented the poster “Bridging the AI Education Gap: A Self-Funded AI Awareness Initiative in Cocoa-Farming Villages of Ghana,” introducing an initiative that teaches basic AI concepts to students in rural Ghana and aims to expand access to AI education in underserved communities.

📍Day 2 @capwic #CAPWIC2026!

🪧Poster session on Cybersecurity, Privacy & Responsible AI! #WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/xIYEm0hIRl
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Susan Zehra, a PhD student and senior lecturer from Old Dominion University’s CS department, presented her poster “Securing Vehicular Ad Hoc Networks (VANETs) Against Cyber Threats,” proposing a decentralized security system to protect vehicle communication networks and showing it can effectively detect and prevent cyber attacks.

Research Shorts: Cybersecurity, Trust & Resilience

Nareman Hamdan from James Madison University chaired the research short session on the “Cybersecurity, Trust & Resilience” track.

📍Day 2 @capwic #CAPWIC2026!

🗯️The Research Short session on Cybersecurity, Trust & Resilience was the last session of the event!#WomenInComputing #VT_IAC @virginia_tech
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

The first presenter, Stephanie Travis from Virginia Tech, presented “Identifying Human Factors in Red Teams for Cyber Exercises.” She focused on making cybersecurity training more realistic by considering how real attackers think and behave. She studied existing works and gathered insights from experts to create a set of human behavioral factors to incorporate in cyber defense exercises in simulations. She found that by improving how red teams simulate attacks, the training can better reflect real-world situations.

💬 Stephanie Travis from @virginia_tech is presenting human factors taxonomy for cyber defense exercises.#Cybersecurity pic.twitter.com/vWffScmxIS
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Next, I presented “Framework for Finding Attribution of Social Media Screenshots.” Sharing screenshots on social media platforms is now common. I pointed out legitimate reasons why people share screenshots, such as to enable cross-platform sharing, to use as evidence for deleted posts etc. Then, I showed how people can create fake tweets easily and share such screenshots on social media platforms. Next, I demonstrated different ways the live web and web archives can be used to find attribution of screenshot content. I emphasized using web archives to find attribution of deleted posts since they cannot otherwise be found on the live web. Lastly, I shared my evaluation results of the automated process of how one can find attribution of a screenshot using the Wayback Machine.

📍Day 2 @capwic #CAPWIC2026!

🗣️ Excited to talk about our research on "Framework for Finding Attribution of Social Media Screenshots" in the Research Short session of Cybersecurity, Trust & Resilience!#WomenInComputing #VT_IAC @virginia_tech @WebSciDL @oducs #SocialMedia pic.twitter.com/teQFvAOfJX
— Tarannum Zaki (@tarannum_zaki) March 30, 2026

Next up, Xinyi Zhang from Virginia Tech presented “From Vulnerable to Resilient: Examining Parent and Teen Perceptions on How to Respond to Unwanted Cybergrooming Advances.” Cybergrooming is a harmful online behavior that can affect teens’ mental health and physical safety. The presenter studied how teens and parents react to different scenarios and identified behaviors that can either increase risk or help protect against harm. By analyzing these responses, she developed patterns of both vulnerable and protective actions to better support teens through education and tools that would encourage safer online behavior.

💬 Xinyi Zhang from @virginia_tech is discussing teen-centered understanding of cybergrooming in thr context of educational programs.#MentalHealth pic.twitter.com/Q7vuGX7tFy
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Yeana Bond from Virginia Tech was the last presenter and discussed improving how metadata-related bugs are detected in Java applications in “Towards Large Language Model-Powered Automation of Detecting Metadata Related Bugs.” Misuse of metadata can cause severe issues in Enterprise Applications written in Java, so her goal was to make debugging metadata problems more efficient with the help of AI. By comparing different AI models, she found that newer models produce more accurate and complete rules.

💬 Yeana Bond from @virginia_tech is demonstrating the potential of LLMs for validating rules for automated metadata debugging tools.#LLMs pic.twitter.com/uSACJhyjM9
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Keynote #3: Human-Centered Automation: A Journey through HCI, AI, and the Future of Robotics

Jessica Zeitz from University of Mary Washington introduced the final keynote speaker Meg Dickey-Kurdziolek, a UX Lead/Senior Staff UX Researcher at Intrinsic.ai. She shared her journey from her PhD at Virginia Tech to her current role at Intrinsic.ai, and how her understanding of human-centered design has changed along the way. She provided valuable insights into the evolving role of UX in the age of AI and robotics. She discussed the challenges of making complex robotic systems more user-friendly. She talked about how Explainable AI (XAI) helps people better understand and trust these systems as AI becomes part of our everyday life. In summary, the speaker highlighted how HCI principles can guide the future of automation and provided useful insights into navigating this rapidly evolving field.

📍Day 2 @capwic #CAPWIC2026!

💬 The last keynote of the session was by Meg Dickey-Kurdziolek from Intrinsic.

She suggested:

*Prioritize systems that increase human competence.
*Don't force humans to match machine requirements.#WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/rWPYMupy6e
— Tarannum Zaki (@tarannum_zaki) March 29, 2026

Closing Remarks and Award Ceremony

The conference concluded with acknowledgments to the sponsors and a vote of thanks to all participants and organizers, followed by an award ceremony. The organizing committee delivered closing remarks and announced that CAPWIC 2027 will be held at James Madison University in Harrisonburg, Virginia. They also introduced the organizing chairs for the upcoming conference. Next, the awards for ‘Best Research Short’, ‘Best Flash Talk’, and ‘Best Poster’ were announced in both graduate and undergraduate categories, along with honorable mentions for each category.

📍Day 2 @capwic #CAPWIC2026!

👏 Closing remarks happening!

The event ended with awards for best presentations:

🏆 Research short
🏆 Flash talk
🏆 Poster

There were awards for honorable mentions too for these categories🏅#WomenInComputing #VT_IAC @virginia_tech pic.twitter.com/1TAlTiM0vD
— Tarannum Zaki (@tarannum_zaki) March 30, 2026

I was delighted to receive the ‘Best Research Short’ award in the graduate category for “Framework for Finding Attribution in Social Media Screenshots.”

The awardees for Best Flash Talk and Best Poster (graduate category) are listed below:

Best Flash Talk – “Benchmarking and Advancing Generative Models for Calorimeter Shower Simulation” by Farzana Yasmin Ahmad from the University of Virginia
Best Poster – “Well-Being in AI-Assisted Software Development” by Fairuz Nawer Meem from George Mason University

🎉 Delighted to share that I won the award for "Best Presentation" in Research Short category for the talk on "Framework for Finding Attribution in Social Media Screenshots" 🏆@WebSciDL @oducs pic.twitter.com/uijuMSoyqb
— Tarannum Zaki (@tarannum_zaki) March 30, 2026

Wrap-up

CAPWIC indeed provides a supportive and encouraging platform for sharing ideas while fostering meaningful opportunities for both personal and professional growth. This was my first time attending the CAPWIC conference in-person. It was a great opportunity for me to connect with researchers, students, and professionals across different areas of computing. I would like to express my sincere gratitude to ODU ACM-W for providing travel support to attend this conference. I also had the wonderful opportunity to stroll through one of the oldest areas in the U.S. – Old Town, Alexandria. I was mesmerized by the brick sidewalks, cobblestone streets, historic townhouses, cherry blossoms in bloom, and the beautiful sunset views along the waterfront. It was refreshing to relax after a full day at the conference.

Previous trip reports for CAPWIC by WS-DL members: 2025, 2015.

---- Tarannum Zaki (@tarannum_zaki)

Artefacto: Finding AI Learning resources for Library Professionals

2026-04-15T10:04:34+00:00

We’ve recently added a new starter curriculum for AI to our Libraryskills.io platform – a space dedicated to highlighting and signposting great, free learning resources for and by library professionals. AI is one of the most talked about topics in libraries right now. And it has particular relevance for library and information professionals for a [...]

Source

Web Science and Digital Libraries (WS-DL) Group at Old Dominion University: 2026-04-13: The Contemplative Sciences Center (CSC) Sensemaking Symposium 2025: Trip Report.

2026-04-14T17:54:39+00:00

Introduction – The CSC Sensemaking Symposium 2025

From October 9–11, 2025, the University of Virginia’s Contemplative Sciences Center (CSC), UVA brought together artists, scientists, scholars, and contemplatives from across the world for the Sensemaking Symposium, a two-and-a-half-day immersion into how humans perceive, interpret, and create meaning in an age defined by complexity.

Hosted in the newly opened Contemplative Commons, the event dissolved traditional boundaries, merging research with lived experience. Through sound installations, musical performances, guided contemplations, and cross-disciplinary conversations, participants entered a space where inquiry became sensory, where knowledge was not only discussed but embodied.

The symposium framed sensemaking as both an inner and collective act, linking neuroscience and mysticism, technology and ritual, sound and silence. Across four thematic sessions, Sensemaking, Hearing, Seeing, and Extrasensory, presenters invited the audience to explore the full range of human perception.

Michael R. Sheehy – Director of Research, Contemplative Sciences Center (UVA)

Dr. Michael R. Sheehy delivered the opening keynote, Contemplative Technologies of Human Sensemaking. As Director of Research at the CSC and Research Associate Professor of Religious Studies, Sheehy bridges lived contemplative traditions, especially Tibetan Buddhism, with modern scientific inquiry. His work spans lucid dreaming, Dzogchen meditation, cognitive illusion, and the cultural ecologies of contemplative practice.

Old Dominion University – Mindfulness and Data Class Participants

The Mindfulness and Data class at Old Dominion University examines how contemplative practices intersect with data-driven ways of understanding the world. Under the guidance of Dr. Nicole Willock, students learn to connect inner awareness with analytical thinking by exploring how attention, emotion, and perception shape the interpretation of information. The course blends reflection, discussion, and hands-on engagement with technologies such as physiological sensors, EEG devices, and digital tracking tools. Through this interdisciplinary approach, students gain insight into how mindfulness can enhance critical thinking, empathy, and more ethical, human-centered uses of data.

As part of their learning journey, students from the class attended the Contemplative Science and Art Conference, where they experienced firsthand how contemplative practice, cutting-edge technology, and creative expression come together. By interacting with tools such as heart-rate monitors, EEG headsets, eye-tracking systems, and immersive art installations, students were able to connect classroom concepts with real-world applications and deepen their understanding of how science can illuminate the inner dimensions of human experience.

	Dr. Nicole Willock is a Professor at Old Dominion University whose teaching bridges mindfulness, religion, and cultural studies. She leads the Mindfulness and Data class and guided her students to the symposium to experience how science, art, and contemplative practice converge.
	Lawrence Obiuwevwi is a doctoral student in Computer Science at Old Dominion University. His research focuses on emotion sensing, physiological signals, and spectrum analysis. He supported the learning experience by helping interpret technologies used at the symposium such as EEG, heart-rate sensors, and eye-tracking systems.
	Cora Morgan is a Communication major whose interests center on storytelling, culture, and mindfulness. The symposium expanded her view of how contemplative science connects inner awareness with social expression.
	Alexis R. Morel is a Psychology major interested in emotional balance and empathy. Her participation deepened her understanding of how the mind and body interact through attention and awareness.
	Araceli Gordus Huizar is majoring in Women’s and Gender Studies with minors in Spanish and Media Studies. She explores identity, culture, and creativity, and found the symposium rich with interdisciplinary insight.
	Dabre Ali is an undergraduate with a growing interest in human-centered technology. The symposium exposed him to tools that illuminate invisible dimensions of human experience, emotion, rhythm, and inner stillness.

Session I – Sensemaking

Opening the symposium, this session explored how human perception, art, and science coalesce into new modes of understanding. Led by eco-artist Wolfgang Buttress, philosopher Jelena Markovic, Buddhist studies scholar James Gentry, and techno-artist David Glowacki, the session wove ecological awareness, embodied knowledge, and the aesthetics of complexity.

Moderated by Devin Zuckerman, the conversation grounded the symposium’s theme: that sensemaking is both analytical and aesthetic, bridging data and devotion, science and spirituality.

Wolfgang Buttress – Eco-Artist, United Kingdom

Dr. Wolfgang Buttress brought an artistic and ecological dimension to Session I: Sensemaking, illuminating how sound, light, and structure can translate the intelligence of the natural world into human experience. A celebrated British sculptor and installation artist, Buttress is internationally known for creating large-scale, data-driven works that merge art, science, and environment. His most renowned installation, The Hive, originally commissioned for the UK Pavilion at Expo 2015 and now a permanent piece at Kew Gardens, uses live sensor data from a real beehive to generate light and sound, allowing viewers to feel the hum of a living colony.

His recent UVA installation, NINFEO, continued this exploration by immersing participants in a responsive landscape of light and resonance. Through collaborations with scientists, architects, and musicians, Buttress transforms empirical data into sensorial art, revealing the hidden rhythms of ecosystems and reminding audiences that perception itself is an act of ecological relationship. His work stands as a meditation on interconnection, an artistic invitation to listen to the world’s subtle frequencies and rediscover the harmony between human sensemaking and the planet’s living pulse.

Jelena Marković – Philosopher, Université Grenoble Alpes, France

Philosopher Dr. Jelena Marković offered a deeply introspective counterpoint to the artistic and scientific discussions, inviting participants to reflect on how thought itself becomes embodied. A scholar whose work bridges philosophy of mind, cognitive science, and affective experience, Marković explores how attention, emotion, and grief transform the self and shape our perception of the world.

Currently a post-doctoral fellow at Université Grenoble Alpes and a member of the Centre for Philosophy of Memory, she examines transformative experiences, moments such as loss or wonder that reorganize our sense of being, and how affect biases attention and meaning-making. Her research also extends into performance and art-based philosophy, using creative forms to investigate cognition and embodiment. At the symposium, her contribution grounded the dialogue in phenomenology and emotional depth, revealing that sensemaking is not only a cognitive act but a lived process through which feeling, memory, and awareness co-construct reality.

James D. Gentry – Buddhist Studies Scholar, Stanford University

Dr. James D. Gentry offered a historical and contemplative lens on how meaning emerges through embodied ritual and sensory practice. An Assistant Professor of Religious Studies at Stanford University and a leading scholar of Tibetan Buddhism, Gentry studies the material, ritual, and visual cultures that shape Buddhist experience. His acclaimed book, Power Objects in Tibetan Buddhism: The Life, Writings, and Legacy of Sokdokpa Lodrö Gyeltsen, explores how objects such as relics, amulets, and ritual implements become vehicles of perception and transformation.

His broader research investigates how sound, sight, and touch function as technologies of enlightenment within Himalayan traditions, and how these sensory frameworks can dialogue with contemporary understandings of consciousness and materiality. At the symposium, Gentry’s reflections bridged ancient contemplative knowledge and modern philosophical inquiry, reminding the audience that sensemaking, whether through ritual or research, is always an embodied and relational act of seeing, hearing, and touching the sacred in everyday life.

David Glowacki – Techno-Artist & Scientist, Intangible Realities Laboratory (Spain)

Dr. David Glowacki expanded the boundaries of perception by merging physics, philosophy, and mystical imagination. Founder of the Intangible Realities Laboratory (IRL), Glowacki is a scientist-artist whose work transforms data into immersive, contemplative experience. With a Ph.D. in molecular physics and a background spanning chemistry, literature, and philosophy, he creates multi-sensory VR installations that invite audiences to feel energy, form, and consciousness as living presences.

At the symposium, Glowacki spoke about his ongoing exploration of Tara, the Buddhist embodiment of compassion and luminous awareness, and how archetypes reveal the human capacity to perceive interconnection beyond the material. Through projects like Isness, which allows participants to merge as glowing fields of light in shared VR space, Glowacki bridges scientific and contemplative traditions, where atoms and awareness coexist in the same luminous field.

Session II – Hearing: Sound & Silence

Sound became the medium through which participants listened to the world anew. JoVia Armstrong drew on her expertise as a percussionist to show how rhythm and reverb shape emotion and meaning. Patrick Finan, a clinical psychologist, described how sonic vibration and music therapy alleviate pain and restore balance to the nervous system. Kythe Heller blended poetry, performance, and theology, while Adam Lobel invited stillness as the acoustic of awareness.

Moderated by sound artist Matthew Burtner, the panel revealed that hearing, whether through silence or resonance, is both a physical and contemplative act, tuning the self to the wider harmonics of life.

JoVia Armstrong – Percussionist, Composer & Assistant Professor of Music, University of Virginia

Dr. JoVia Armstrong transformed the symposium space into a living instrument, reminding us that sound is not merely heard, it is felt. A percussionist, composer, and sound artist from Detroit, Armstrong blends Afro-diasporic rhythmic traditions with modern electronic experimentation to explore the emotional and spatial dimensions of listening.

At the symposium, she spoke about the importance of reverb in musical composition, not just as an acoustic effect but as a metaphor for resonance, echo, and memory. In her words, reverb gives sound a body; it situates the listener within space and time, allowing emotion to linger like breath in a room.

Currently an Assistant Professor of Music at the University of Virginia, Armstrong holds a Ph.D. in Integrated Composition, Improvisation, and Technology from UC Irvine. Her performance ensemble, Eunoia Society, experiments with drones, loops, and multichannel environments to create immersive sonic meditations.

Through her work, she bridges rhythm and reflection, tradition and technology, inviting audiences to experience sound as a contemplative process of becoming aware of one’s own presence in the auditory world.

Patrick H. Finan – Clinical Pain Psychologist & Professor of Anesthesiology, University of Virginia

Dr. Patrick Finan invited listeners to consider sound as medicine, a bridge between psychology, physiology, and the inner landscape of pain. A clinical pain psychologist and Harold Carron Professor of Anesthesiology at the University of Virginia, Finan’s work explores how sleep, emotion, and reward systems shape the experience of chronic pain.

At the symposium, he discussed the healing potential of sound and music, describing how rhythm and resonance can modulate emotional states and neural activity, providing moments of relief and restoration.

Drawing from research in his lab, including fMRI, sensory testing, and ecological momentary assessment, Finan explained that music’s ability to soothe pain is grounded in psychophysiological synchronization: the body literally entrains to patterns of calm.

His talk framed listening as an act of empathy and self-regulation, suggesting that the future of pain management may rely not only on medication but on cultivating deep, attentive sonic relationships with one’s own body.

Kythe Heller – Interdisciplinary Artist, Poet & Scholar, Harvard University

Dr. Kythe Heller wove poetry, philosophy, and performance into a meditation on the spiritual and sensory dimensions of listening. An interdisciplinary artist and Doctor of Theology from Harvard University, Heller’s work bridges creative practice and contemplative inquiry, exploring how sound, silence, and language become conduits for transformation.

At the symposium, she reflected on the voice as a vehicle of revelation, tracing how resonance and vibration carry meaning beyond words, invoking both the mystical and the material. As founder and director of Vision Lab at Harvard Divinity School, Heller convenes artists, scientists, and contemplatives to explore imagination as a force that reshapes consciousness.

Her poetry collection Firebird and multimedia performances investigate illumination, grief, and transfiguration. Through her presence, Heller invited participants to experience silence not as emptiness but as a vibrant medium, a threshold where self and world meet in shared reverberation.

Adam Lobel – Contemplative Teacher, Ecophilosopher & Founder of 4F Regeneration

Dr. Adam Lobel invited participants to listen beyond the human, to tune into the soundscape of the Earth itself. A contemplative teacher, ecophilosopher, and founder of 4F Regeneration, Lobel works at the intersection of Buddhist practice, ecological awareness, and collective transformation.

He spoke about sound as a bridge between consciousness and the living world, encouraging listeners to recognize hearing as both an ethical and ecological act. Drawing on his background in Buddhist philosophy and decades of contemplative teaching, Lobel suggested that awareness practices can heal the rift between human perception and planetary systems.

His teachings blend meditation, ritual, and ecological activism, creating spaces for embodied reflection. In Charlottesville, his words rang like a dharma bell, reminding participants that every sound, from wind to breath to silence itself, is a pulse in the shared heartbeat of life.

Kelsey Johnson – Astronomer & Professor of Astronomy, University of Virginia

Dr. Kelsey Johnson guided participants to look outward, and inward, through the lens of the night sky. A renowned astronomer at the University of Virginia and founder of Dark Skies, Bright Kids!, she studies the birth of galaxies and the formation of stars hidden within cosmic dust.

Johnson spoke about the loss of the natural night sky due to light pollution, reminding the audience that the glow of cities is dimming humanity’s oldest connection to the cosmos. For millennia, humans oriented their stories, rhythms, and sense of humility by the stars.

She argued that regaining dark skies is both an ecological and contemplative act, an invitation to rediscover our place in the vastness of space. When we lose the stars, she reflected, we risk losing sight of our own smallness and wonder. Her talk blended scientific insight with existential reverence, making the night sky a mirror for meaning and fragility.

Andrew Holecek – Contemplative Teacher, Author & Scholar of Dream Yoga

Dr. Andrew Holecek invited participants to journey beyond ordinary perception, to explore the “luminous darkness” of the mind itself. A renowned teacher of Tibetan Buddhist meditation and lucid dreaming, Holecek has spent decades studying how awareness continues through waking, dreaming, and dying.

He spoke about the transformative power of darkness, drawing on Tibetan dark retreat practices where total darkness reveals the inner light of consciousness. He described how the night, both literal and psychological, can become a field for insight rather than fear, showing that seeing is not only optical but spiritual.

Author of works including Dream Yoga: Illuminating Your Life Through Lucid Dreaming and the Tibetan Yogas of Sleep, Holecek emphasized that cultivating awareness in darkness dissolves boundaries between seer and seen. His reflections reminded the audience that light and darkness are partners in perception, and that embracing the unseen helps us see more clearly within.

Jesse Fleming – Media Artist & Assistant Professor of Emerging Media Arts, University of Nebraska–Lincoln

Dr. Jesse Fleming explored how technology, light, and presence shape perception. A filmmaker, media artist, and Assistant Professor at the University of Nebraska–Lincoln’s Carson Center for Emerging Media Arts, his work bridges consciousness studies, design, and immersive art.

Fleming spoke about how mediated seeing can become a contemplative practice, reflecting on how screens, reflections, and moving images influence attention and empathy. Drawing on projects such as The Shared Individual and Nuclei, he demonstrated how immersive environments can expand awareness rather than fragment it. His research asks what happens when the media stops entertaining and starts awakening, when pixels, photons, and human perception synchronize to reveal the subtle boundary between observer and observed. Fleming reframed “seeing” as participation in a living network of light, bodies, and consciousness.

Session IV – Extrasensory

The final session expanded perception beyond the five senses, merging science, spirituality, and technology. Mikey Siegel introduced bio-sensor experiences that transform heartbeats and breath into shared light and sound fields. Eve Ekman illuminated the emotional body as a sensory organ guiding compassion and resilience. Michael Lifshitz traced the neuroscience of hypnosis, meditation, and psychedelics to show how consciousness continually remakes reality. Oludamini Ogunnaike offered a luminous account of Sufi and Islamic practices in West Africa, where chant, rhythm, and beauty serve as portals to divine knowledge. Moderated by Casey Forgues, the discussion synthesized art, science, and spirituality into one realization: sensemaking begins where the measurable meets the mystical.

Mikey Siegel – Technologist & “Consciousness Hacker”, Stanford University

Dr. Mikey Siegel brought frontier-thinking to the table, showing how technology and collective physiology become tools of sense-making. A former robotics engineer (MIT Media Lab) now based at Stanford University and working with his initiative BioFluent Technologies, Siegel designs immersive systems (such as his renowned platform GroupFlow) that measure participants’ heart-rate and breath and convert these into shared audio-visual experiences.

At the symposium he spoke about how sense-making isn’t just a solo act of cognition, but a field phenomenon, a resonant space where bodies, devices, sounds and attention interweave. He urged us to ask not only what our technologies do, but who they enable us to become.

Eve Ekman – Contemplative Social Scientist & Emotion Researcher, University of California, Berkeley

Dr. Eve Ekman turned attention inward, guiding participants to consider emotion itself as a sensory organ, a compass for meaning and human connection. A contemplative social scientist and Senior Fellow at the Greater Good Science Center at UC Berkeley, Ekman’s research explores emotional awareness, empathy, and resilience.

She spoke about how emotions shape perception, emphasizing that sensemaking is not limited to intellect or sensation but is deeply informed by the body’s internal signals. Drawing from her work on the Atlas of Emotions with the Dalai Lama and her Cultivating Emotional Balance program, she illustrated how mindfulness and compassion training help individuals transform reactivity into clarity.

Her reflections revealed that emotional literacy is a contemplative technology, one that allows people to feel more deeply, connect more authentically, and perceive the subtle vibrations of the human heart. Ekman reminded participants that the future of awareness depends not only on sharper tools of observation but on gentler capacities for feeling.

Michael Lifshitz – Neuroscientist & Assistant Professor of Psychiatry, McGill University

Dr. Michael Lifshitz bridged neuroscience, anthropology, and contemplative practice to examine how the human mind constructs , and transcends , ordinary perception. An Assistant Professor of Psychiatry at McGill University and Director of the Psychedelics and Contemplation Lab, Lifshitz investigates how meditation, hypnosis, and psychedelics alter consciousness and the sense of self.

At the symposium, he spoke about how non-ordinary states of awareness reshape the boundaries of the senses, describing them as experiments in human possibility. Drawing from neuroimaging and ethnographic research, he explored how spiritual and contemplative experiences can transform the brain’s perception of agency and embodiment.

His talk emphasized that what we call “extrasensory” may not be supernatural at all, but an expanded form of sensemaking that includes the body, culture, and consciousness in continuous dialogue. Through this lens, Lifshitz offered a scientific and deeply human reminder: to understand the mind, we must study not just what it perceives, but how it learns to see itself.

Oludamini Ogunnaike – Associate Professor of African Religious Thought & Democracy, University of Virginia

Dr. Oludamini Ogunnaike explored how sensory experience and spiritual knowledge fuse in the Sufi and Islamic traditions of West Africa, inviting participants to listen for the hidden frequencies of sacred sound, poetry, and devotion. At the University of Virginia, he teaches African religious traditions, Islamic philosophy and art, and the intellectual history of Sufism and Ifá.

His research examines the aesthetic, philosophical, and sensory dimensions of West African Islamic and indigenous traditions, particularly how devotional recitations, mystical poetry, and ritual practices function as forms of sense-making. At the symposium he described how the chants of the Tijāniyya order, the rhythms of madīḥ poetry, and the oracular Ifá tradition reveal the senses as conduits of knowledge, not just passive receptors.

Through works such as Deep Knowledge (2020) and Poetry in Praise of Prophetic Perfection (2020), he reframes sense-making as a poetic, embodied, and spiritual act, one in which the boundaries between listener, liturgy, and divine presence dissolve.

Conclusion – Integrating the Senses, Integrating the Self

The Contemplative Sciences Symposium revealed the power of interdisciplinary inquiry, where artists, scientists, philosophers, physicians, and contemplative practitioners came together to examine how humans perceive, interpret, and make meaning. Across sessions on hearing, seeing, extrasensory awareness, and the nature of sensemaking itself, the symposium showed that understanding the world requires more than intellect alone; it requires the full participation of the senses, the body, and the imagination. This gathering demonstrated how deeply interconnected the contemplative, scientific, and creative disciplines truly are, and how each contributes a vital perspective to the study of awareness and human experience.

A significant part of this success is also reflected in the structure and philosophy of Dr. Willock’s Mindfulness and Data course, which seamlessly integrates contemplative practice with physiological measurement, emotional awareness, and data-driven inquiry. In her classroom, students learn to read heart rate, breath, and affective signals not merely as metrics, but as reflections of lived, embodied processes. By guiding students to unite mindfulness with analytic rigor, Dr. Nicole creates a learning environment in which theory becomes experience and data becomes self-knowledge. Her approach shows that education can be contemplative, scientific, and personal all at once, inviting students to think critically, feel deeply, and cultivate attention as a tool for understanding.

The participation of her students in the Contemplative Sciences Symposium further exemplifies this integration. Engaging directly with leading scholars, artists, and contemplative researchers allowed them to witness interdisciplinary collaboration in action and to situate their own learning within a broader landscape of inquiry. Through both classroom practice and conference immersion, students experienced firsthand how mindfulness, physiology, art, culture, and neuroscience converge to expand human understanding. This synergy, between curriculum and community, between inner practice and academic exploration, highlights the transformative potential of contemplative education and the essential role it plays in shaping thoughtful, reflective, and compassionate learners.

Lawrence Supervisors’ Special Thanks:

I would like to express my sincere gratitude to my supervisors, Dr. Erika Frydenlund, Research Associate Professor at Old Dominion University and a member of the Storymodelers Lab; Dr. Krzysztof J. Rechowicz, Assistant Professor at Old Dominion University and a member of the Storymodelers Lab and the Virginia Digital Maritime Center (VDMC); and Dr. Sampath Jayarathna, Associate Professor at Old Dominion University and a member of the Web Science and Digital Libraries Research Group and the NIRDSLab, for the continued opportunities they have afforded me to be part of impactful research and meaningful academic endeavors. Their guidance, support, and mentorship have been invaluable to my growth.

About the Author:
Lawrence Obiuwevwi is a Ph.D. student in the Department of Computer Science, a graduate research assistant with The Center for Secure and Intelligent Critical Systems (CSICS), and a proud student member The Web Science and Digital Libraries (WS-DL) Research Group, and NirdsLab at Old Dominion University.

Lawrence Obiuwevwi
Graduate Research Assistant
Virginia Modeling, Analysis, & Simulation Center
Department of Computer Science
Old Dominion University, Norfolk, VA 23529
Email: lobiu001@odu.edu
Web : lawobiu.com

Digital Library Federation: Digital Storytelling in Practice: A New Session Format for the DLF Forum

2026-04-14T12:00:24+00:00

Digital Storytelling in Practice: A New Session Format for the DLF Forum

Team DLF is introducing a new session format in the Call for Proposals for the 2026 Virtual DLF Forum: Digital Storytelling Presentations. This format is designed to deepen collaboration, center relationships, and create space for shared learning across roles, institutions, and communities.

The new 40-minute Digital Storytelling (DS) Presentation format is designed as an interactive session that highlights digital storytelling projects developed through collaborative partnerships. These DS Presentations center on installation-inspired projects, such as exhibits, platforms, or collections, that offer immersive, experiential engagement for participants. We encourage presenters to incorporate demonstrations whenever possible to help attendees engage more fully with the tools, platforms, or storytelling approaches being shared.

Rather than focusing solely on a single presenter or project overview, the format should feature a minimum of two (2) presenters and no more than three (3) presenters. For example, a digital librarian or archivist might pair with a community partner, student, artist, or scholar whose work is represented in, or inspired by, the digital project. Together, presenters will explore not only the final product but also the collaborative process, relationships, and ideas that shaped the work, to show attendees how this work might be imagined, adapted, and implemented within their own institutions.

Presentations will emphasize the broader significance of digitization, why access matters, how collections are used, and the impact beyond the institution. Examples of proposals might include:

Archive to Art: A digital archivist and artist show how digitized protest materials inspired a multimedia installation, emphasizing workflow and creative impact. Example: Women’s March on Washington and Atlanta March for Social Justice and Women Collection (January 21, 2017), Women and Gender Collections, Georgia State University Library Digital Collections

Community Memory in Motion: A librarian and historian built a neighborhood digital archive through collaboration, now used in schools and local programs. Example: Folded Map Project

Teaching with Data: A librarian and student used a digitized collection to create a data visualization project, linking the technical process to student research. Example: Students Turn College Fight Songs into Award-Winning Data Visualization | News | Northwestern Engineering

This format is included in the 2026 Call for Proposals, and we look forward to seeing how presenters bring collaborative digital storytelling to the Forum. If you’d like to talk through your idea or learn more about the format, please email us at forum@diglib.org.

The post Digital Storytelling in Practice: A New Session Format for the DLF Forum appeared first on DLF.

Digital Library Federation: Call for Proposals: 2026 Virtual DLF Forum

2026-04-14T12:00:06+00:00

CLIR’s Digital Library Federation (DLF) invites proposals for the virtual 2026 DLF Forum, to be held online October 14-15, 2026. Learn more about who we are and who attends the DLF Forum.

Please note: This Call for Proposals (CFP) is for the October virtual event. There is no in-person event in 2026. We are committed to making this online conference accessible to all through consistent use of captioning in all sessions and the provision of accessible presentation materials, screen-reader-friendly documents, and clear communication of accommodation options. For accessibility related questions or concerns, please contact forum@diglib.org.

The submission deadline is Monday, May 11, at 11:59 pm ET

We invite proposals for live virtual presentations on all topics related to digital libraries, encompassing case studies, “show and fails,” practical applications, methods, projects, ethics, research, and learning in any area, such as:

Collections & Stewardship: Digitization, digital preservation, digital asset management systems (DAMS), born-digital materials, and format conversions.
Community & Advocacy: Partnerships, community archives, outreach, and professional advocacy.
Digital Research & Pedagogy: Digital humanities, scholarship, music, art, creative expression, and digital pedagogy.
Ethics, Justice, & Society: Race and technology, accessibility, AI/Machine Learning, copyright, and environmental sustainability.
Infrastructure: Platforms, workflows, project management, and assessment.

This list of content topics is intended as a starting point and is not exhaustive; we welcome additional ideas and approaches that align with the spirit of the Forum.

Session Formats

All sessions will take place live in a meeting-style or webinar-style Zoom room, and breakout rooms will be available upon request for all formats except lightning talks. Sessions are invited in the following lengths and formats:

- 90-minute Workshops: Guided training sessions on a specific tool, technique, workflow, or concept. Up to five (5) facilitators are allowed per submission.
- 50-minute Working Sessions: Open sessions for community organizers, creative problem solvers, and existing or prospective DLF working groups to begin or get feedback on in-progress projects, collaborate on addressing challenges, and discuss thought-provoking questions. Up to five (5) facilitators are allowed per submission.
- 40-minute Panels: Discussions of up to five (5) presenters on a unified topic, with an emphasis on community discussion. Proposals with diverse and inclusive speaker involvement will be favored by the committee. Panels will be slotted into 50-minute sessions, leaving a minimum of 10 minutes for Q&A and discussion at the end of each session.
- 40-minute Presentations: A single topic or project presented by up to three (3) presenters. Presentations will be slotted into 50-minute sessions, leaving a minimum of 10 minutes for Q&A and discussion at the end of each session.

NEW! 40-minute Digital Storytelling (DS) Presentations: Interactive sessions highlighting digital storytelling projects—such as exhibits, platforms, or collections— developed through collaborative partnerships that offer immersive, experiential engagement. They should feature a minimum of two (2) and no more than three (3) presenters in conversation, such as a digital librarian or archivist paired with a community partner, student, artist, or scholar whose work is represented in, or inspired by, the digital project. Demos are encouraged. DS Presentations will be slotted into 50-minute sessions, leaving a minimum of 10 minutes for Q&A and discussion at the end of each session. Read more about this new format here.
5-minute Lightning Talks: High-energy talks on any topic held in succession in a single session, presented by up to two (2) presenters. There is no formal Q&A for lightning talks, but we encourage presenters to share contact information with attendees for follow-up conversations after the session.

Proposal Requirements

Proposal title and submission format.
Author information: full names, organizational affiliations, and email addresses for all presenters and authors.
Brief abstract – limited to 50 words. This abstract will appear in Community Voting and in the conference program.
Full proposal – limited to 250 words for all formats except for panels and workshops, which are limited to 500 words. This full proposal will only be seen by reviewers and the Program Committee.
Five keywords for your proposal.
Breakout room request – there will be an option in the submission form to indicate a request for breakout rooms.
Workshops Only: Learning objectives (limited to 50 words; brief, clear statements about what attendees will be able to do as a result of taking your proposed workshop); technology needed; participant proficiency level; how your workshop will be interactive.
Session materials (notes, documents, slides, handouts, etc.) will be shared under a CC BY 4.0 license, which allows for sharing and adaptation of content with appropriate credit and an indication of any changes made. We will continue to invite presenters to deposit these materials in the Zenodo.org open-access repository, where the DLF community archives DLF Forum materials under this license. Presenters must agree in the submission form to share their materials under these terms.

Submissions and Evaluation

Based on community feedback and the work of our Program Committee, we welcome submissions geared toward a practitioner audience that:

Clearly engage with DLF’s mission;
Activate and inspire participants to think, make, and do;
Engage people from different backgrounds, experience levels, and disciplines; and/or
Include clear takeaways that participants can integrate and implement in their own work.

All submissions will be peer-reviewed. Reviewers will use this rubric to rate each proposal based on the values listed above. They may also recommend the proposal for a different format. Broader DLF community input will also be solicited through an open community voting process, which will help inform the Program Committee’s final decisions.

We especially welcome proposals from individuals who bring diverse professional and life experiences to the conference, including those from underrepresented or historically excluded racial, ethnic, or religious backgrounds, immigrants, veterans, those with disabilities, and people of all sexual orientations or gender identities. As we have done in the past, the Program Committee will prioritize submissions from individuals who identify as Black, Indigenous, and People of Color (BIPOC), individuals working at Historically Black Colleges and Universities (HBCUs), Tribal Colleges and Universities (TCUs), Hispanic Serving Institutions (HSIs), Minority Serving Institutions (MSis) and other libraries, archives, museums, and organizations that center BIPOC to promote inclusivity to the greatest extent possible. Self-identification options will be provided in the proposal submission form, but are not required.

Schedule

Call for Proposals opens: Tuesday, April 14
Call for Proposals closes: Monday, May 11
Notification of final decisions: Week of June 11
Program released: Week of June 23

Please feel free to reach out with any questions: forum@diglib.org.

FAQ

What is the DLF Forum?

DLF programs stretch year-round, but we are perhaps best known for our signature event, the DLF Forum.

The DLF Forum welcomes digital library, archives, and museum practitioners from member institutions and beyond—for whom it serves as a meeting place, marketplace, and congress. As a meeting place, the DLF Forum provides an opportunity for our working groups and community members to conduct their business and present their work. As a marketplace, the Forum provides an opportunity for community members to share experiences and practices with one another and support a broader level of information sharing among professional staff. As a congress, the Forum provides an opportunity for the DLF to continually review and assess its programs and its progress with input from the community at large.

Here, the DLF community celebrates successes, learns from mistakes, sets grassroots agendas, and organizes for action. The Forum is governed by the DLF’s Code of Conduct. All Forum, in-person and online events, and community participants are expected to uphold a harassment-free, inclusive environment. The Code prohibits bullying, discrimination, and harmful behavior of any kind, requires respectful, constructive engagement and adherence to established safety protocols, and includes options for reporting harassment.

Generally, who attends the DLF Forum?

The DLF Forum attendees are a multi-disciplinary cross-sector community of people who work in the digital library, museum, archives, and cultural heritage fields, from librarians, project managers, curators, technologists, and developers to administrators and service providers. The Forum welcomes practitioners from academic, art and cultural heritage, and non-profit organizations, government agencies, and more. They come from all over the country and world and represent all levels of professional experience. Forum attendees are inquisitive, engaged, and action-oriented with a focus on learning new skills and solving problems together. When offered in a virtual format, the DLF Forum may reach a wider and larger audience than in-person events.

Who should submit a proposal?

We encourage proposals from:

DLF members and non-members;
Regulars and newcomers;
Digital library practitioners from all sectors (higher education, museums and cultural heritage, public libraries, archives, etc.) and those in adjacent fields such as institutional research and educational technology;
Students, early- and mid-career professionals, and senior staff alike.

Can you help me find a co-presenter?

Looking for co-presenters on a particular topic? Try using our 2026 DLF Forum Unofficial Program Sessions and Connections spreadsheet for connecting with other prospective presenters. Note that the Program Committee and CLIR+DLF Staff do not monitor the document and it is not part of the official submission process.

How is my proposal evaluated?

All submissions will be peer reviewed. Reviewers will use this rubric to rate each proposal. They may also recommend the proposal for a different format. Broader DLF community input will also be solicited through an open community voting process, which will inform the Program Committee’s final decisions.

What makes a successful proposal? Can I see successful proposals from previous years?

Based on community feedback and the work of our Program Committee, we welcome submissions geared toward a practitioner audience that:

Clearly engage with DLF’s mission;
Activate and inspire participants to think, make, and do;
Engage people from different backgrounds, experience levels, and disciplines; and/or
Include clear takeaways that participants can integrate and implement in their own work.

We strongly encourage prospective presenters to review our rubric and past DLF Forum programs (from 2025 and 2024 in-person & virtual) to understand what makes a successful DLF Forum proposal. Strong proposals will demonstrate how presenters intend to design their proposed sessions to be interactive, inclusive, and action-oriented, and will also outline clear learning objectives. We especially welcome proposals from individuals who bring diverse professional and life experiences to the conference, including those from underrepresented or historically excluded racial, ethnic, or religious backgrounds, immigrants, veterans, those with disabilities, and people of all sexual orientations or gender identities. As we have done in the past, the Program Committee will prioritize submissions from individuals who identify as Black, Indigenous, and People of Color (BIPOC), individuals working at Historically Black Colleges and Universities (HBCUs), Tribal Colleges and Universities (TCUs), and other libraries, archives, museums, and organizations that center BIPOC to promote inclusivity to the greatest extent possible. Self-identification options will be provided in the proposal submission form, but are not required.

What is the author limit? What is the presenter limit?

Each session type has a maximum number of presenters per submission:

90-minute Workshops: Up to 5 facilitators
50-minute Working Sessions: Up to 5 facilitators
40-minute Panels: Up to 5 presenters
40-minute Presentations: Up to 3 presenters
40-minute Digital Storytelling Presentations: Up to 3 presenters
5-minute Lightning Talks: Up to 2 presenters

There is no limit to the number of non-presenting authors listed on a proposal.

The post Call for Proposals: 2026 Virtual DLF Forum appeared first on DLF.

LibraryThing (Thingology): Come join the TinyCat 10th Birthday Hunt!

2026-04-13T20:44:00+00:00

We’re hosting a special TinyCat Birthday Treasure Hunt over on LibraryThing! We’ve got ten clues, one for each year. We’ve scattered a collection of birthday banners around the two sites, and it’s up to you to find them all.

Come brag about your clowder of tiny cats (and get hints) on Talk.
Decipher the clues and visit the corresponding pages in LibraryThing or TinyCat to find a banner.
Each clue points to a specific page. Remember, some banners will be hidden in LibraryThing and some in TinyCat!
You have until 11:59 pm EDT on Thursday, April 30th to find all the TinyCats.

Win prizes:

Any member who finds at least two birthday banners will be awarded a TinyCat banner badge.
Members who find all 10 birthday banners will be entered into a drawing for one of five sets of TinyCat and LibraryThing swag. We’ll announce winners at the end of the hunt.

P.S. Thanks to conceptDawg for the gray catbird illustration!

Ed Summers: Weekly Bookmarks

2026-04-12T04:00:00+00:00

These are some things I’ve wandered across on the web this week.

🔖 AI Whistleblower: We Are Being Gaslit By AI Companies, They’re Hiding The Truth! - Karen Hao

The truth about Sam Altman. AI Critic Karen Hao reveals what 90 OpenAI employees told her.

Karen Hao is an AI expert, award-winning investigative journalist, and former reporter for The Wall Street Journal covering American and Chinese tech companies. She is also co-host of the podcast The Interface and freelances for publications like More Perfect Union and The Atlantic. Her latest book is the bestselling ‘EMPIRE OF AI: Inside The Reckless Race For Total Domination.’

🔖 Introduction to Compilers and Language Design

This is a free online textbook: you are welcome to access the chapter PDFs directly below. If you prefer to hold a real book, you can also purchase a hardcover or paperback below. The textbook and materials have been developed by Prof. Douglas Thain as part of the CSE 40243 compilers class at the University of Notre Dame. Join our mailing list to receive occasional announcements of new editions and other updates.

🔖 Inside Claude Code With Its Creator Boris Cherny

A somewhat bizarre interview with the creator of Claude Code, where he talks about the origins of the tool, and how its current development fits in with Anthropic’s business plans – which seem pretty vague other than taking over the world.

🔖 Mitchell Hashimoto’s new way of writing code

Mitchell Hashimoto on building HashiCorp, navigating the cloud giants, and how AI agents have transformed his day-to-day engineering workflow.

🔖 pi-mono contributing guide

Some open source projects that accept AI contributions are moving to a model where PRs need to reference an issue that has been marked approved by an existing maintainer with a lgtm comment. This then triggers a Github Action that adds the user to the .github/APPROVED_CONTRIBUTORS file. Then when a PR comes in, it isn’t immediately closed.

https://newsletter.pragmaticengineer.com/p/mitchell-hashimoto

🔖 badlogic / pi-mono

AI agent toolkit: coding agent CLI, unified LLM API, TUI & web UI libraries, Slack bot, vLLM pods.

(apparently Claude Code was built with this?)

🔖 The Coral Bones

The Coral Bones is a tale of three women from different times in Earth’s history, each of whom has a special relationship with the Great Barrier Reef. Judith is the daughter of a 19th Century English sea captain and is desperate to study the natural world, just like the famous Mr. Darwin. Hana is a Japanese-Australia scientist from the present day, studying the dying reef. And Telma is a descendant of refugees in a near future Australia where most forms of animal life except humans are functionally extinct.

🔖 In Ascension

In Ascension is a 2023 novel by Martin MacInnes, published in the UK by Atlantic Books and in the US by Grove Atlantic.[1] It is published or forthcoming in ten languages. The novel tells the story of Leigh, a young girl who grows up in the Netherlands amid the specter of climate change and eventually becomes a marine scientist exploring ocean trenches and investigating an anomaly at the edge of the Solar System.

🔖 Artemis II Mission Dashboard

A dashboard for what is going on right now for the mission and space weather, and hand livestream.

🔖 Papers, Please: The toll of age verification laws on digital sex work

“The only point [of these laws] is to restrict access to content,” Riana Pfefferkorn, an attorney and policy fellow at Stanford University’s Institute for Human-Centered Artificial Intelligence, told me. “I think that the ubiquity of [age verification] lately has warped people’s views of what online safety means, so that now everything is just like, ‘Why don’t we just do [age verification]? Won’t that fix it?’” she continued. She was alluding to the use of AI to generate adult content, as well as the trend of users on X requesting that the platform’s AI assistant, Grok, non-consensually undress photos of potentially underage girls. In response to outcry, X chose to paywall access to its AI tools

🔖 Content Neutrality for Kids: Intermediate Scrutiny for Social Media Age-Verification Laws

The First Amendment imposes a high, but not insurmountable, hurdle for states to overcome in regulating minors’ social media use. By focusing on specific features that lead to harmful effects on minors, states can craft content-neutral laws that will merit only intermediate scrutiny. The solutions to the LinkedIn Problem proposed above — naming platforms directly under TikTok’s revival of the “special characteristics” standard or regulating specific harmful features without reference to content — are the two likeliest ways for states to have their laws upheld in court. Like California, states must be creative and flexible as they respond to a rapidly developing legal doctrine. If “[s]ocial media is a cancer on our society,”213 then seeking a constitutional cure is crucial even if current efforts “dwell only on the suffering of children.”214

🔖 The machines are fine. I’m worried about us.

for someone who doesn’t yet have that intuition, the grunt work is the work. The boring parts and the important parts are tangled together in a way that you can’t separate in advance

🔖 On The Enshittification of Audre Lorde: “The Master’s Tools” in Tech

This is not an argument to reject the enshittification analysis. It is an argument to extend it. A decolonial critique of technology is not simply “the internet was always bad.” It is rather: the conditions that made the internet harmful to specific communities were never peripheral to its design; they’re an integral part of it. And any politics that aims to restore something like the pre-enshittification internet without reckoning with those conditions is doomed to reproduce them.

Web Science and Digital Libraries (WS-DL) Group at Old Dominion University: 2026-04-11: PostGuard: The Hackathon-Winning AI That Stops Career-Ending Posts

2026-04-11T14:21:26+00:00

From March 23rd to the 31st, the Computer Science Graduate Society (CSGS) at Old Dominion University hosted their Spring 2026 Hackathon. The competition brought together teams across master's and PhD categories to tackle different research topics, mainly in artificial intelligence (AI). Our team, the Attention Bros (Sandeep Kalari and Dominik Soós), chose to compete in Track 6: Privacy-Preserving AI, alongside four other great teams in the PhD category.

We are grateful to announce that we won first place in the PhD category with our project, PostGuard!

This was a fast-paced challenge completed over a single week. Despite the time constraints, we successfully engineered a novel architecture that balances AI utility. In this blog post, we provide an overview of the problem we tackled, the Privacy Paradox and existing methods, our system architecture; and the mathematically proven findings that secured our victory.

For more details, you can explore our Github repository containing the code, detailed report, and the dataset used in our analysis.

Online Comments Have Lasting Consequences

Social media platforms serve as both a public forum and a digital newsstand. We started by looking at the problem: online comments can cause irreversible, real-world damage. In the heat of the moment, you post something, and before moderation can catch it, someone takes a screenshot. People lose their jobs over 280 characters posted online.

Current moderation systems are 100% reactive, so they only act after the fact. We wanted to build a preventative system where it warns you before you hit send.

The Privacy Paradox and Existing methods

To give users a specific and actionable warning about how a post violates their employee's policies, the system needs to know their personal context, like their job role and employer. However, collecting and processing that data creates a massive surveillance and privacy risk.

When we looked at how existing research handles this problem, we found a significant gap. To prevent someone from posting something that will ruin their career, you have three standard options, all of which fail:

Content moderation is reactive. It doesn't warn the user; it just punishes them after the fact, and it's also not user-specific.
Differential Privacy works great for aggregate data but is useless for individual, consequence-based warnings.
Text Anonymization frameworks like RUPTA are great at removing personally identifiable information (PII) from text, but they strip away the exact context the LLM needs to generate a personalized warning.

We needed a system that acted pre-posting, user-specific, and privacy-aware. That's why we built PostGuard.

Approach	Reactive?	Pre-posting?	User-specific?	Privacy-aware?
Content Moderation	✅	❌	❌	❌
Differential Privacy	❌	❌	❌	✅
Text Anonymization (RUPTA)	❌	❌	➖ Partial	✅
PostGuard (Ours)	❌	✅	✅	✅

Building a Dataset

To rigorously test our system, we couldn't just use standard benchmarks, so we built a dataset grounded in reality. We spent the first phase of the hackathon collecting a custom dataset:

15 Real Incident Cases: We pulled verified, real-world firings that were covered by major outlets.
20-Article Vector Corpus: We embedded 15 signal articles and intentionally injected 5 noise articles to rigorously test our retrieval precision.
Synthetic Personas: We generated 15 synthetic users with escalating post histories spanning from 2024 to 2026, mapping them one-to-one with real corporate policies.

The Architecture

PostGuard intercepts risky posts before the user hits submission. To do this without leaking the user's data to the web, we built a four-stage pipeline:

Risk Extraction: We use a lightweight LLM (Gemini Flash) to quickly extract risk factors from the draft and generate a targeted search query.
RAG layer: We use an embedding model to search our custom vector database for relevant corporate policies and real-world firing precedents.
Warning Generation: A secondary LLM (Gemini Pro) synthesizes the retrieved precedents and generates a customized, user-facing warning.
RUPTA Evaluation: Finally, we run a dual-evaluation loop. A P-Evaluator scores the re-identification risk of the data we just processed, and a U-Evaluator scores the utility of the generated warning.

Figure 1. System architecture detailing the four stages of evaluation: (1) Initial comment ingestion, (2) Contextual policy retrieval via RAG, (3) Severity classification, and (4) Generation of the intent-preserving warning

Three Privacy Modes

The core of our privacy-preserving approach is user control. The system operates in three modes that dictate what data is forwarded through the pipeline.

Mode	Data Sent to System	Privacy	Utility
Anonymous	Comment text only. No role, no employer, no history.	High — poster nearly unidentifiable	Lower — generic warnings
Contextual	Comment + platform + job role.	Medium — role narrows the field	Medium — role-specific warnings
Full Profile	Comment + role + employer + recent history.	Low — nearly identifiable	High — employer-policy specific warnings

Evaluation Results

The evaluation of our moderation and warning system demonstrates a highly effective balance between accuracy, user intent preservation, and privacy. To understand the system's true performance, we analyzed it across four core dimensions: Privacy vs. Utility, RAG Retrieval Accuracy, Severity Classification, and Warning Quality. Here is a breakdown of the metrics we used and why they matter.

Privacy vs. Utility

Protecting user identity is just as important as providing accurate warnings. By calculating the Relative Utility Threat (RUT) score, adapted from Soonseok Kim's 2025 MDPI Electronics paper, "Quantitative Metrics for Balancing Privacy and Utility in Pseudonymized Big Data", we proved mathematically that Contextual mode (RUT: 0.824) delivers higher AI accuracy than Full Profile Mode, while exposing only 46% of the relative privacy risk.

Mode	RUT Score	Utility	Re-id Risk	Interpretation
Anonymous	0.908	88	0.05	Excellent — high utility, almost no re-id risk.
Contextual	0.824	94	0.35	Best balance — recommended deployment threshold.
Full Profile	0.652	92	0.75	Utility gain does not justify the massive privacy cost.

The RUT framework validates that feeding the LLM maximum personal data ("Full Profile") yields diminishing returns by transforming various risk and utility metrics into a unified, probabilistic scale. Contextual mode sits right at the optimal deployment threshold, giving the AI just enough context, like the job role and platform, to generate highly specific warnings without sacrificing anonymity.

RAG Retrieval Accuracy

To evaluate RAG Retrieval Accuracy, the system was tested on 15 real-world incident cases against a mixed corpus. The results demonstrate highly effective document sourcing, achieving a Hit Rate@1 score of 0.80, meaning that the correct article ranked first in 12 out of the 15 cases. Furthermore, the system also achieved perfect retrieval with a Hit Rate@3 score of 1.00, ensuring the relevant articles was always surfaced within the top three results.

Metric	Score	What it means
Hit Rate@1	0.80	Correct article ranked first in 12/15 cases.
Hit Rate@3	1.00	Correct article always in top 3 — perfect retrieval.
Mean Reciprocal Rank	0.90	Average rank position is very high.

This strong performance is reinforced by a Mean Reciprocal Rank of 0.90, which confirms that the average rank position of the correct information remains consistently high across all queries.

Severity Classification

The system prioritizes user trust and accuracy in Severity Classification, which measures its binary classification performance for detecting high and critical violations. It achieved a perfect Precision score of 1.000, guaranteeing zero false alarms so the system never wrongly warns a safe comment.

Metric	Score	Interpretation
Precision	1.000	Zero false alarms — the system never wrongly warns a safe comment.
Recall	0.533	7 cases under-classified — the system is intentionally conservative.
F1 Score	0.696	Overall classification quality.

The overall classification quality is represented by an F1 Score of 0.696. The Recall score of 0.533 reflects 7 cases that were under-classified. However, this is a deliberate design choice to make sure the system remains conservative rather than over-restrictive.

Warning Quality

Traditional metrics like exact-match or BLEU scores fail to capture the nuance of rewritten text. for this reason, we used an "LLM-as-Judge" framework to score the qualitative aspects of the AI's output on a 5-point scale. This allowed us to measure subjective dimensions like Relevance, Policy Accuracy, Rewrite Safety, and Prevention Impact at scale.

Dimension	Mean Score	What it measures
Relevance	4.67	Does the warning correctly identify the actual violation?
Policy Accuracy	4.67	Does it cite the correct policy or law for this specific case?
Rewrite Safety	4.67	Does the rewrite preserve intent while removing the risk?
Prevention Impact	4.67	Would this warning likely have prevented the real firing?
Overall	4.67	Holistic quality score

The overall mean score of 4.67/5 confirms the system acts as a helpful, accurate coach that preserves the user's original intent while effectively neutralizing the career risk. We found that the system successfully rewrites drafts to preserve their intent while removing career risks.

Looking Forward

The internet doesn't have to be a trap door. With PostGuard, we proved that we can give users specific and potentially career-saving warnings without turning AI tools into surveillance machines.

We are incredibly grateful to the CSGS organizers for putting together such a challenging and rewarding event. Earning first place in the PhD category was the peak of a long, exhausting, and incredibly fun week of research.

Thanks for reading, and watch what you comment!

~Dominik Soós (@DomSoos)

Library | Ruth Kitchin Tillman: Open Refine: Blanking Down Only Within Records

2026-04-11T00:00:00+00:00

This is the first in what will be a series of posts deriving from my most recent use of OpenRefine. This post is primarily for intermediate or advanced OpenRefine users, so I won’t be going over fundamentals. Beginners who are already familiar with the power of “Blank down” and “Transform” should also be ok.

One of the great things about OpenRefine is that you can take a spreadsheet with a repeating key field, apply the “Blank down” function on that column, and unlock a “record” experience. I find this really helpful when I need to combine MARC data with item data, a 1:many relationship. I can facet the holding item library to a particular campus, for example, and see entire records, not just the item row.

While only the key needs to be blanked down to create the record view, I can improve my experience working with the data by blanking down other repeating fields so I’m only seeing one instance of any given field.

The problem is that while keys don’t repeat, other data may. If I’ve got a column for notes and I’m working on a set of related records, multiple adjacent records may have the same note text. Blanking down means that the note text is only left in the first record of the sequence, until there’s a record where that field is genuinely blank or which has a different text.

I’ve previously handled this problem by counting how many rows were impacted by my initial blank down and noting how many were impacted by blank downs on each subsequent column. If the number is higher, I undo my action and simply leave the field duplicated in each row of the record. But I don’t like it.

The Solution

While working on a massive data review project this spring, I became frustrated with repeating data which I couldn’t blank down. I decided to search around online to see if others have found ways to handle it. I found the solution in a 10-year-old thread on the OpenRefine Google Group:

Once you’ve turned your rows into records, apply the following cell transform to each column you want to blank down:

value + " - " + row.record.index

Now, perform the “Blank down” operation.

Get your original content back with a second cell transform:

value.replace(/ - \d+$/,'')

The (Brief) Explanation

It’s important to note here that you must have already performed the initial “Blank down” operation on your column with keys to turn your rows into records. Otherwise, each row will be treated as its own record. Once you have records, though, they’ll be treated as records by this transformation whether you’re currently viewing the data as rows or records.

The first transformation uses the simple string join GREL syntax to join the original value, a " - “, and that row’s record index.

Something like:

Collection Name
Pennsylvania German broadsides and Fraktur
Pennsylvania German broadsides and Fraktur
Pennsylvania German broadsides and Fraktur
Pennsylvania German broadsides and Fraktur
Pennsylvania German broadsides and Fraktur
Pennsylvania German broadsides and Fraktur
Pennsylvania German broadsides and Fraktur
Pennsylvania German broadsides and Fraktur
Pennsylvania German broadsides and Fraktur

becomes:

Collection Name
Pennsylvania German broadsides and Fraktur - 0
Pennsylvania German broadsides and Fraktur - 0
Pennsylvania German broadsides and Fraktur - 0
Pennsylvania German broadsides and Fraktur - 0
Pennsylvania German broadsides and Fraktur - 1
Pennsylvania German broadsides and Fraktur - 1
Pennsylvania German broadsides and Fraktur - 2
Pennsylvania German broadsides and Fraktur - 2
Pennsylvania German broadsides and Fraktur - 2

and when I apply the blank down function, I get:

Collection Name
Pennsylvania German broadsides and Fraktur - 0



Pennsylvania German broadsides and Fraktur - 1

Pennsylvania German broadsides and Fraktur - 2

(This is just one column view, there’s an assumed leftmost column with the record IDs.)

The second transformation is a simple regex replace, looking for the delimiter (” - “) and a string of one or more numbers up to the end of the cell value. Because we’ve put a right-anchor on it, it should only match the addition we made, even if the string happens to contain its own " - 92951” (unless you’ve turned on repeating).

This simple pair of transformations and the ability to blank down without losing data has really improved my experience of working with the large data exports.

It’s a bit cumbersome to repeat them each time, so stay tuned for my next blog post: “I can’t believe I’ve been using OpenRefine for a dozen years and only just learned how easy it is to repeat a set of operations.”

Zotero: Zotero 9

2026-04-10T20:46:17+00:00

We’re excited to announce Zotero 9, which introduces a major new way to engage with your documents, along with a host of other improvements to the research workflow.

Coming less than three months after Zotero 8, Zotero 9 is the first major update since we announced a faster release cycle for Zotero, and it represents our commitment to getting stable features into the hands of users more quickly.

Read Aloud

Read Aloud reads your documents to you in high-quality, natural-sounding voices. It works on PDFs, EPUBs, and webpage snapshots.

To get started, just click the headphones button in Zotero’s reader toolbar.

As you’re listening, you can skip forward or backward by paragraph or by sentence (Option/Alt-click or Option/Alt-left/right), and you can start reading from a particular point by clicking in the left margin or by right-clicking and choosing “Read Aloud from Here”.

An “Annotate Sentence” button — or H or U on your keyboard — will automatically highlight or underline the last sentence you heard, and you can use shortcut keys to quickly move, expand, or delete the new annotation.

Your last reading position is saved and synced between devices, so you can pick up where you left off on any device.

Read Aloud requires an internet connection and a Zotero account for high-quality voices, which we’re calling Zotero Voices. If you’d like to use Read Aloud offline, you can still use the text-to-speech voices available on your system, but the quality will be significantly degraded.

We offer two tiers of Zotero Voices: Standard and Premium. Standard voices are generated on Zotero servers, with unlimited minutes for Zotero Storage subscribers and 2 hours/month for free accounts. Premium voices are the highest-quality voices available, processed by external text-to-speech providers — they make fewer mistakes, sound more natural, support many more languages, and can handle multilingual text. Individual Zotero Storage subscribers will receive up to 2 hours of Premium usage (varying by voice) each month, and free and institutional accounts will also receive a small quota in order to try the voices out. Initially, all subscribers can request additional Premium minutes for free. In the near future, we’ll provide more details on monthly allocations and options for adding additional minutes going forward.

Read Aloud is currently available only in the desktop app, but it’ll be coming to the iOS and Android apps soon.

Recently Read

A new Recently Read collection at the top of the collections list in each library shows items with attachments you’ve recently read, most recent first. Opening an attachment or changing pages will bump the item to the top of the list.

The collection includes a Last Read column, which you can also add to other views, and “Attachment Last Read” is available as an Advanced Search condition.

The last-read time syncs between devices, so you can quickly find a file on another device that you’ve read elsewhere.

Insert Annotations Directly into Word Processor Documents

The word processor plugins now feature a new Add Annotation button that lets you insert one or more annotations directly into your document, with active Zotero citations that automatically generate bibliography entries.

Previously, you could add annotations to Zotero notes and insert notes into your document with active citations, but it’s no longer necessary to create an intermediate note if you prefer to work directly in your word processor.

Add Annotation opens a new mode in the citation dialog that expands attachments to show individual annotations. You can browse or search for annotations, choose one or more, and then add them to your document, along with active citations and any comments you added. Image and ink annotations are inserted as images.

“Added By” and “Modified By” for Group Libraries

You can now add “Added By” and “Modified By” columns to the items list in group libraries, letting you see and sort by the people who created and last updated items.

These fields also show in the metadata list along with Date Added/Modified.

Per-Group File Renaming Settings

Group admins can now configure file-renaming settings for each group library, ensuring consistent filenames for all group members.

Performance Improvements

We’ve made some major improvements to Zotero’s performance, including reducing startup memory usage by 20% and drastically reducing disk access and network requests during file syncing in some situations.

On macOS, Zotero now uses a feature of the modern Apple filesystem to avoid additional disk-space usage when copying files. This includes the automatic daily backups Zotero makes of its database, potentially saving hundreds of megabytes or gigabytes of local disk space.

Web-Based Login

You now log in to your Zotero account via the browser instead of entering credentials in the app. This allows you to use a password manager to auto-fill credentials and will enable two-factor authentication (currently in beta), greatly increasing the security of your Zotero account.

Other Improvements

See the changelog for the full list of changes.

Get Zotero 9

If you’re already running Zotero, you can upgrade from within Zotero by going to Help → “Check for Updates…”.

Don’t yet have Zotero? Download Zotero 9 now.

Open Knowledge Foundation: Knowledge as Critical Digital Infrastructure: A Call to Action for a Resilient Future

2026-04-10T10:32:55+00:00

You can also read and share this story in Spanish and Portuguese. Knowledge is the foundation of every modern society, underpinning democracy, driving innovation, and strengthening our collective culture. In the digital age, this bedrock is critical digital infrastructure (CDI), the essential software, standards, data systems, and information that provide public functions upon which society depends. By recognizing...

The post Knowledge as Critical Digital Infrastructure: A Call to Action for a Resilient Future first appeared on Open Knowledge Blog.

LibraryThing (Thingology): Happy 10th Birthday TinyCat!

2026-04-09T18:51:51+00:00

Ten years ago we created TinyCat, a catalog solution for small libraries.

The idea was simple: Thousands of small libraries were already using LibraryThing, but it was too much—too many doors to the larger world of discovery and social interaction. So we cut all that out, and we made everything about finding books. We added the circulation and administration features small libraries need. We made the simple and intuitive user interface we wished that big, public libraries had.

Growth and Improvements

From about 500 libraries in our first year, we have grown to more than 3,500 today—from schools to religious communities, from museums and LGBTQ centers, to a growing number of small public libraries. Another 32,516 LibraryThing members use TinyCat as a second way to search their personal libraries.

The last decade has brought TinyCat to countries across the globe, from the USA to Ireland, Singapore, Peru, Egypt, and South Africa to name a few. We recently added translation options, so more people can now use TinyCat in more languages.

There are many small libraries who still don’t know about TinyCat, and we can use your help spreading the word. If you know anyone working or volunteering with a small library, send them our way. Folks can sign up or learn more about TinyCat at https://www.librarycat.org.

TinyCat Survey

Share your thoughts about TinyCat in our new survey!

This survey is short and every question is optional. You can save your progress and come back any time. While the survey is not anonymous, we encourage you to be honest and candid — all your feedback helps us improve!

This is a great chance to share your experiences as an admin, feedback from your patrons, and ideas you have for making TinyCat even better. All current and previous TinyCat members are welcome to participate.

New Feature: Restrict Catalog Access

TinyCat libraries now have the option of requiring a login to view their catalog. If enabled, all visitors will be prompted to enter their credentials before seeing the homepage or searching the catalog.

To enable this for your TinyCat library, go to your Patron Accounts settings and switch on Restrict Catalog Access.

Enter the Giveaway!

Enter our TinyCat 10th Birthday Giveaway for a chance to win a special TinyCat bookmark and TinyCat stickers!

How to enter:

Share one of your favorite or most popular items from your TinyCat library on social media. Alternatively, you’re welcome to post in the TinyCat group on LibraryThing.
Include a screenshot or link to the item’s catalog page so we can see it like your patrons do.
Tag TinyCat in the post so we can find your post and enter you in the giveaway.

(Not just books! Game libraries can share a game, movie libraries a film, etc.)

Where to tag us:

Facebook: TinyCat
Bluesky: @librarycat.org
X: @TinyCat_lib
LibraryThing Talk Group (link coming soon)

Deadline:

We will randomly select 10 winners on May 8, 2026. Each will receive a special TinyCat bookmark and stickers. We will contact the winners for mailing details.

Store Sale

All TinyCat merchandise and barcode scanners, including stickers, pins, and coasters, are on sale now through May 4. Barcode scanners are only $5 and TinyCat pins are only $2! See all the discounts at the LibraryThing Store.

Profile Badges

TinyCat libraries over 1, 2, 5, and 10 years respectively will receive new badges in LibraryThing. Go ahead, brag!

Thank you for reading and for celebrating with us!

Mita Williams: If large language model use is ascendant and unavoidable, why is there so little discussion about the importance of vocabulary?

2026-04-09T16:18:09+00:00

There is a moon in the sky called the moon. Oh, you wanted a better description?

Web Science and Digital Libraries (WS-DL) Group at Old Dominion University: 2026-03-17: The Disintegration Loops: Generational Loss in Web Archives

2026-04-08T18:48:35+00:00

The Disintegration Loops: Generational Loss in Web Archives

Michael L. Nelson

As part of the Internet Archive's Information Stewardship Forum (March 18–20, 2026), I decided to use my five minute lightning talk to raise the issue of generational loss in web archives. Or more directly, making copies of copies (...of copies…) – something that web archives currently do not do well. My title is based on William Basinski's four volume release "The Disintegration Loops", in which he played the audio tapes of "found sounds", recorded decades earlier, in loops, with the whole process lasting over an hour. The effect is hauntingly beautiful, with each loop slightly degrading the magnetic tape, resulting in a generational loss. The degradation of each loop is right on the edge of the just-noticeable difference, until the entire track is reduced to just a shadow of its former self.

I first discussed this topic in my 2019 CNI closing keynote (slide 88), where I introduced the inability of web archives to archive other web archives as part of the larger issue of web archive interoperability. Let's begin with walking through the example of archiving a tweet (which we already know to be challenging!). The original tweet is still on the live web, even though the UI has undergone many revisions since when it was originally tweeted in 2018.

https://twitter.com/phonedude_mln/status/990054945457147904

(screen shot from 2026-03-17)

I archived that tweet to the Internet Archive's Wayback Machine in 2018 (screen shot from 2019):

https://web.archive.org/web/20180501125952/https://twitter.com/phonedude_mln/status/990054945457147904

I then archived the Wayback Machine's copy of the tweet to archive.today in 2019 (screen shot from 2019):

https://archive.ph/PaKx6

Note that archive.today is aware that the page comes from the Wayback Machine but the original host is twitter.com, and it maintains both the original Memento-Datetime (20180501125952) as well as its own Memento-Datetime (20190407023141). I then archived archive.today's memento to perma.cc in 2019 (screen shot from 2019):

https://perma.cc/3HMS-TB59

Finally, I archived the perma.cc memento back to the Wayback Machine in 2019 (screen shot from 2019):

https://web.archive.org/web/20190407024654/https://perma.cc/3HMS-TB59

Although the loss occurs in discrete chunks, it is reminiscent of Basinski's Disintegration Loops, with information lost at each step, and the final version being a mere shadow of the original. In 2019, this was not universally recognized as a problem, since archiving the playback interface of other web archives was not considered a problem to itself. The "right" solution, of course, is to share the WARC files (or WAC, or HAR, or…) out-of-band and let the other web archives replay from the same source files. But this is rarely possible: for a variety of reasons web archives typically do not share the original WARC files, and in the case of archive.today, might not even store the original source files (and instead, likely only store the radically transformed pages).

More importantly, it is sometimes useful to archive a particular web archive's replay of a page, which itself must be archived, because it changes through time. For example, memento #3 (the perma.cc memento of archive.today's memento) is now different; this is a screen shot from 2026:

2026 replay of https://perma.cc/3HMS-TB59

Surely the source files themselves have not changed, and the difference is due to improvements in pywb, which is under constant development. perma.cc's replay of the 2019 page in 2019 is different from the replay from 2026, which implies that it could be different still in the future. But we can not currently archive without generational loss of perma.cc's replay of that page to, say, the Wayback Machine. The fact that screen shots – which are rife with their own potential for abuse (cf. HT 2025, arXiv 2022) – are the only mechanism to document these replay differences underscores the web archive interoperability problem.

I chose the topic of generational loss for my slot at the Information Stewardship Forum because recent events have introduced a new use case for archiving the replay of web archives. Wikipedia recently announced it was blacklisting archive.today because its editors discovered that webmaster at archive.today was using its captcha to direct a DDoS attack against a blog owned by someone that webmaster had a dispute with (the blogger had posted a lengthy investigation of the identity of webmaster), and, for our discussion more disturbingly, had edited the content of an archived page to include the name of the blogger where it would not otherwise be. The Wikipedia discussion page is hard to follow, in part because the editors are discussing how to archive the replay of an archived page. For one example, they show how the archive.today replay now has been changed back to have "Comment as: Nora ████" (middle of the image):

But the replay alteration from archive.today in question is archived at megalodon.jp to show that the name "Nora ████" was replaced with the name of the blogger that had earned webmaster's ire, "Jani Patokallio". And yes, megalodon.jp's replay of archive.today's memento is that bad (at least in my browser, it is shrunk down impossibly small), so I used the dev tools to find the string in question.

https://megalodon.jp/2026-0219-1509-14/https://archive.is:443/2021.05.30-173350/http://www.maskofzion.com/2012/04/jewish-at-root-iraqs-destruction-hell.html

Another Wikipedian archived (using yet another archive, ghostarchive.org) a google.com SERP to show that archive.today has reverted from "Jani Patokallio" back to "Nora ████".

What does changing "Nora" to "Jani" (and then changing it back again) accomplish? I'm not sure; this appears to be just a petty response to an ongoing dispute. But the implication is profound: this is the first known example of a major web archive purposefully and maliciously altering its contents, something that we knew was possible but had not yet experienced.

We have long known that replay can change through time (cf. PLOS One 2023) due to the replay engine (the Wayback Machine, Open Wayback, pywb, etc.) evolving, but these changes were engineering results and the replay mostly improved over time. But now we have seen web archives maliciously alter (and then revert) the replay, and we need a more standard and interoperable way to archive archival replay. Not just to prove that a web archive did alter its replay, but also to prove that an archive did not alter its replay. Out-of-band sharing of WARC files is the gold standard, but for a variety of reasons this is unlikely to happen. We must be able to use web archives to verify and validate web archives. We explored a heavyweight design for this a few years ago (JCDL 2019), but it should be revisited in light of developments like WACZ.

–Michael

ht to Herbert Van de Sompel for introducing me to "The Disintegration Loops" many years ago.

2026-03-25: The original Google/Blogger name ("Nora") has been anonymized.

Open Knowledge Foundation: An Honest Reflection on the Integration of LLMs into Open Data Portals

2026-04-07T18:02:09+00:00

‘We don’t trust AI,’ claims OKFN's Tech Lead in this text, in which he offers his sharp views on one of today's most pressing challenges

The post An Honest Reflection on the Integration of LLMs into Open Data Portals first appeared on Open Knowledge Blog.

LibraryThing (Thingology): Author Interview: Shelley Noble

2026-04-07T12:45:45+00:00

LibraryThing is pleased to sit down this month with best-selling author Shelley Noble, whose many novels run the gamut from historical fiction to mystery to contemporary women’s fiction. A former professional dancer, Noble toured with Twyla Tharp Dance and American Ballroom Theater, and has worked as a choreographer for film and theater productions. She earned her BFA and MFA at the University of Utah, and taught at California State University in Fresno. A former president of Sisters-in-Crime, Noble is a member of Mystery Writers of America, Romance Writers of America, and Liberty States Fiction Writers, and currently lives in New Jersey. Her newest novel, The Sisters of Book Row, was published by William Morrow in March 2026 and tells the story of three sisters and bookstore proprietors who confront the Comstock laws in 1915 Manhattan. Noble sat down with Abigail this month to discuss the book.

How did the story idea for The Sisters of Book Row first come to you? Were you drawn to the thought of writing about bookstores and booksellers, or perhaps about the Comstock laws?

I’ve had Comstock in the back of my mind for a while, a perfect villain, a vicious zealot, one who I particularly despise. So when my editor suggested I write a book about books, guess who came to mind. And because I write about Manhattan, I knew the perfect place in which to set the story, Book Row, once the mecca of rare and used book buyers from around the world. And like a magnet, this germ of an idea began collecting bits and pieces. An article about the current Cohen sisters of the Argosy Book Store inspired me to create the Applebaum sisters, and Sisters was born.

Tell us about the Comstock laws. What were they, and what effect did they have on the world of books and booksellers, as well as the wider American society of that time?

Anthony Comstock moved to New York in the early 1870s and was appointed special agent to The Society for the Suppression of Vice and the U.S. Post Office to prevent pornography from being sent through the mail. He was given the power to search, seize, arrest and fine, the monies of which he received half. His activities quickly spread to all facets of life, and as his power grew, his ideas of what was “obscene, lewd, or lascivious,” changed, sometimes from week to week. Later in his career, his extreme and outlandish views made him a laughing stock, ridiculed by the newspapers, and dismissed by the courts. The Post Office fired him, but he refused to leave. The NYSPV replaced him, but again he ignored them and continued on his crusade. The Comstock Act, enacted in 1873, included a ban on contraception and was written by Comstock himself. It was never repealed, but Roe v. Wade relegated it to being a zombie law. Unfortunately states had adopted the original law for their own use. And today we see it being used to prevent birth control information, or any reproductive health measures from all women. A zealot, who is said to have destroyed 15 tons of books and four million pictures and other materials, who hated women and died ridiculed and despised, and yet he has managed to rear his ugly head again today.

Your story is set on Book Row, a district in lower Manhattan that contained over three dozen bookstores at its height. Did you have to do any research about the history of the area, and what were some interesting things you learned? If you could visit any bookstore from that period, which would it be? (Disclosure: I worked for the Strand bookstore—the sole survivor of Book Row—for many years).

I used to hang out at the Strand all the time. Many years ago. It was a solace and an adventure away from the chaos of the city and my profession as a young dancer. I hope my Sisters of Book Row can come to life for readers of today. I did loads of research, I always do. It’s one of my favorite parts of writing historical fiction. There’s a lovely book titled Book Row by Marvin Mondlin and Roy Meador. It didn’t have as much information on my particular period 1915 as I hoped, but it was fascinating to read about the continuation of this community, especially post 1930.

Once I get an overview of my time and place and characters, I like to depend mainly on primary sources, newspapers, anecdotes, letters. That way I know what they know, feel what they feel, and try to leave my historical outsider’s knowledge at the door. I mostly learned the neighborhood in bits and pieces since the area has built up so much since then.

Some oddities and coincidences: The Argosy, owned by Louis Cohen, the father of the Cohen sisters who added their inspiration to this story had his own run in with “Comstockery” in the 1930s. When the city began digging the new subway, customers couldn’t get around construction, and many stores had to move uptown, then moved back in when it was completed. After I developed and had lived with my three Applebaum sisters and the Arcadia for weeks and several chapters, I learned that there was actually a Mr. Applebaum who had a bookshop in the Row, named Arcadia. Did I read about it and forgot while it became ingrained in my subconscious? Or was it really a coincidence? I was too attached to my own Applebaums to change their names, so I mentioned the existence of two families in my Author Notes.

Sometimes a story is like a jigsaw puzzle, learning a phrase, a sentence about the inhabitants. The two booksellers, who were constantly arguing, gave me an image that led to the daily morning conversations around the newsstand. They might have argued and complained, but they were neighbors and they were ready to take up a collection to bail one of their own out of jail when Comstock was on the prowl.

The book world has been rocked in recent years by an upsurge of attempts at censorship and book suppression. I chronicle some of that in the Freedom of Expression column of our monthly State of the Thing newsletter. What can your story tell us about our situation today, in this respect?

For our modern selves, I wish The Sisters of Book Row and their withstanding the attacks of what they loved most was so outside of our experience, so unbelievable, that readers might say. “Oh, that would never happen here.” But unfortunately we see it happening throughout our country by those who, like Comstock, denounce books they’ve never even read and bully those who only want to share knowledge. Their attacks sometimes seem so diffuse and widespread that we might think it will never affect us. It will, but I have to believe that we’re more experienced, more aware of the rotten core of the book banning movement, and that if we keep up a constant resistance, we will prevail.

Tell us a little bit about your writing process. Do you have a particular routine—a schedule you keep, or a place you like to write? You write in a number of different genres, does your story-building process differ, depending on the genre?

I do have a routine though it has changed over the years and books. When I wrote two books a year, I had a tighter schedule. Now that I’m writing one historical I can linger in the research, jump down a rabbit hole or two. And I find that writing of the past, I’ve changed from being an early morning writer, to a late night writer. There’s something about the dark and the quiet that I find conducive to delving into the past. Of course the nearer I get to deadline, the more daytime writing I have to do. I have a home office where I write all my books. Each genre requires a different energy and attitude. The contemporaries don’t require as much deep dive research, so I can begin writing sooner than with the historicals. No matter the genre, I depend on a storyboard to keep everything on track. Not a computer screen board but a big gridded Lucite board on the wall with color coded post-its for characters and plot points that can be moved around as the story develops.

What comes next for you? Are there any new books you’re currently working on?

I’m currently working on a story that takes place in 1870 Long Branch, New Jersey, where President Grant has his summer capital and a young woman aspiring to become a lawyer confronts the changes and the scandals that threaten the quiet seaside town she calls home.

Tell us about your library. What’s on your own shelves?

Lots of history books, mainly early 20th century New York, late 19th American theatre books when the Rialto was Union Square. Dickens, Austen, Mary Stewart. A rotation of women’s historical fiction. Eastern religion. Mystery and science fiction. I’m a pretty eclectic reader.

What have you been reading lately, and what would you recommend to other readers?

This fall I decided to go on a rereading spree. I started with Fahrenheit 451 followed by 1984 right before the holidays. Yes, they are still as scary as when I read them in school. After that, I immediately pulled out my favorite chapters of The Pickwick Papers. Now I’m re-rereading The Hobbit, and reading A Founding Mother** about Abigail Adams, by Stephanie Dray and Laura Kamoie. I highly recommend all of these.

**Stay tuned for our interview with Stephanie Dray and Laura Kamoie this coming July, in honor of America’s 250th birthday!

Cynthia Ng: Yet another resume (cv) writing guide

2026-04-06T03:53:51+00:00

Many years ago, I wrote a tips on job applications post, because I had been seeing a lot of poorly formatted and written applications. While good formatting and tailoring to job postings are important, the content is obviously important as well. Again, there is so much literature out there, but what I’ve seen recently has … Continue reading

Nick Ruest: Infrastructures of Listening: The ManoWhisper Podcast Analysis Pipeline

2026-04-06T00:00:00+00:00

ANGEL:	Greetings, Prophet;
	The Great Work begins;
	The Messenger has arrived.

Planet Code4Lib

Journal of Web Librarianship: A Streamlined AI-Powered Workflow for Enriching Library Newspaper Archives

Coral Sheldon-Hess: AI policies and informed consent

Contents of this post

Why there’s no incentive for management to require AI usage:

Why people deserve informed consent before using these products:

An additional AI policy suggestion:

And last, but in a just world, not least:

The benefits are usually negligible, but often nonexistent and sometimes even negative

Now

The future

Harms to individual users

AI usage damages critical thinking and encourages cognitive surrender

AI products encourage dependence and decrease self-confidence

AI usage hinders skill formation and learning on novel tasks

AI usage erodes expertise and damages performance even on familiar tasks

AI usage amplifies individuals’ biases and flaws in judgment, in ways that are invisible to them

Sending uncorrected “workslop” to someone else is always inappropriate

AI outputs can be hard to parse

AI outputs must be understood to be corrected (and they always need to be corrected)

Ethics and the environment

I want to shout out to people, books, and articles that helped organize and clarify my thinking, even if I didn’t cite them explicitly here

And why not share my TBR (to-be-read) list, while we’re here

A final note:

Xe Iaso: Maybe you shouldn't install new software for a bit

Open Knowledge Foundation: The Neurotech Booth at RightsCon That Never Happened

HangingTogether: No edge case: Understanding AI opportunities through Arabic metadata workflows

LibraryThing (Thingology): Author Interview: Katie Holt

David Rosenthal: The Permissionless Catch-22

White Hats

Catch-22

Black Hats

Conclusion

Open Knowledge Foundation: Brazil and OKFN partner to enable citizens to ‘talk’ to national data

Library | Ruth Kitchin Tillman: OpenRefine: API Basics and Repeat Operations

API Basics

Authentication

Performing Operations With the API

Evaluating Efficiency

Ed Summers: Weekly Bookmarks

🔖 Soundcamp / Reveil

🔖 endetymes

🔖 Gray Area

🔖 RubyLLM

🔖 Contributor Poker and Zig’s AI Ban

🔖 We did the math on AI’s energy footprint. Here’s the story you haven’t heard.

🔖 Before GitHub

🔖 Do I belong in tech anymore? On quitting, the spread of AI, and the loss of an ideal.

🔖 Getting my daily news from a dot matrix printer

🔖 Claude-powered AI coding agent deletes entire company database in 9 seconds — backups zapped, after Cursor tool powered by Anthropic’s Claude goes rogue

🔖 WARC Extensions: From idea to stable specification

🔖 Twenty Years Later Twitter, Transformation, and Platform Legacies

🔖 Speaker notes from Resisting Big Tech Empires

🔖 Safety-I and Safety-II: The Past and Future of Safety Management

🔖 Non serviam

🔖 What are embeddings?

🔖 Talk Talk - Live at Montreux 1986

🔖 The West Forgot How to Make Things. Now It’s Forgetting How to Code

🔖 There Will Be a Scientific Theory of Deep Learning

Web Science and Digital Libraries (WS-DL) Group at Old Dominion University: 2026-05-02: Evolution of Language-Guided Control in UAV Systems

LibraryThing (Thingology): May 2026 Early Reviewers Batch Is Live!

Digital Library Federation: DLF Digest: May 2026

This month’s news

This month’s open DLF group meetings:

Get Involved / Connect with Us

Artefacto: Five for Friday – AI policy examples for libraries

Dan Cohen: A Framework for Books and AI in the Public Interest

Open Knowledge Foundation: Learning together, building together – OKFN Newsletter April 2026

Open Knowledge Foundation: Join Open Knowledge at RightsCon 2026, Zambia

David Rosenthal: Dormant Digital Assets

HangingTogether: New OCLC Research Report: The Library Beyond the Library

Ed Summers: Weekly Bookmarks

🔖 Talk Talk - Live at Montreux 1986

🔖 The West Forgot How to Make Things. Now It’s Forgetting How to Code

🔖 There Will Be a Scientific Theory of Deep Learning

🔖 EHRAG: Bridging Semantic Gaps in Lightweight GraphRAG via Hybrid Hypergraph Construction and Retrieval

🔖 The Night Manager (British TV series)

🔖 What Will It Take to Get A.I. Out of Schools?

🔖 iocaine

🔖 AI as a Fascist Artifact

Update 22^nd April 2026

Update 24^th April 2026